Monthly Estimates of U.S. Cross-Border Securities Positions^*

Carol C. Bertaut and Ralph W. Tryon

NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International Finance Discussion Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors. Recent IFDPs are available on the Web at http://www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from the Social Science Research Network electronic library at http://www.ssrn.com/.

Abstract:

This paper reports monthly estimates of U.S. cross-border securities positions obtained by combining the (now) annual TIC surveys with monthly transactions data adjusted for various differences in the two reporting standards. Our approach is similar to that of Thomas, Warnock, and Wongswan (2004), but in addition to having a somewhat larger dataset we are able to make some simplifications to the numerical procedure used and we incorporate additional adjustments to the transactions data. This paper describes the procedure used and presents the monthly results. In addition, we discuss how the procedure can be extended to extrapolate holdings estimates beyond the most recent survey values. We focus primarily on U.S. liabilities to foreign holders, because more data is available than for U.S. claims, but we show how our methodology can be applied to U.S. claims as well. We also provide some guidance on how the changes in estimated holdings can be decomposed into flows, valuation changes, and other factors. Time series of estimates of holdings, by country, are available for download.

Keywords: International investment position, treasury international capital, cross-border securities Holdings.

JEL classification: C80, F30.

1 Introduction

Cross-border holdings of securities between the United States and the rest of the world are growing in size and importance. As of end-December 2006, U.S. residents held about $5.6 trillion in foreign stocks and bonds, compared with holdings of $2.1 trillion five years earlier, while foreign residents held about $8 trillion in U.S. long-term securities, more than double their holdings in 2001. Periodic surveys of holdings provide our most accurate and detailed information on cross-border securities holdings, but these surveys have several disadvantages. First, the surveys are relatively infrequent: until recently, surveys were as much as five years apart. They are now conducted annually, but a higher frequency time series of positions is desired by market participants and policy analysts. Second, survey results are available only with a considerable lag: it takes about 8-9 months to get preliminary results and 10-12 months to get final results, after the nominal date of the survey. We would like a reliable way to base current estimates of holdings on the latest monthly securities transactions data, which are available with a lag of only 45 days, approximately. Finally, the surveys alone do not provide a basis for decomposing changes from one survey to the next into net transactions, valuation effects, and other adjustments, all of which would be helpful in analyzing the data. By estimating monthly positions using monthly transactions and valuation adjustments, we can estimate the desired decomposition as well.

We work with three sets of cross-border securities data collected by the Treasury International Capital (TIC) system. First, foreign holdings of U.S. securities are measured in the comprehensive surveys of U.S. liabilities to foreigners. These data are available by country of holder, by security type (Treasury bonds, agency bonds, corporate bonds, and equities), and by type of holder (official or private) for nine different dates: December 1984, December 1989, December 1994, March 2000, June 2002, June 2003, June 2004, June 2005, and June 2006.¹ There are about 80 countries in the sample.² Second, on the claims side, surveys of U.S. holdings of foreign securities were taken in March 1994, December 1997, December 2001, December 2003, December 2004, December 2005, and most recently for December 2006. Claims data are available by country of issuance for U.S. holdings of foreign bonds and foreign equities. We combine these periodic survey data with the third set of data, monthly transactions data on cross-border purchases and sales of U.S. Treasury, agency, corporate bonds, U.S. equities, and foreign stocks and bonds (the TIC S data).³ Although the most recent liabilities and claims surveys provide considerable detail on the types of securities held (e.g. currency of issue, "straight" debt versus asset-backed, zero-coupon, or convertible debt, public versus private issuer, common stock versus preferred stock or mutual funds), the monthly transactions data limit our analysis to the broad characterizations listed.

The difficulties involved in making monthly estimates, particularly those caused by financial center transactions bias, have been discussed in a number of other papers (See Warnock and Mason (2001), Griever, Lee, and Warnock (2001), and Warnock and Cleaver (2002).) Thomas, Warnock, and Wongswan (2004) propose a methodology to generate monthly position estimates by country that use adjusted monthly transactions and are consistent with the reported survey positions. Our approach, discussed in the following sections, is similar to that of Thomas, Warnock, and Wongswan (hereafter TWW), but differs in a number of details.

2 The discrepancy between survey positions and transactions-based positions estimates

There is a substantial discrepancy between the reported survey positions and position estimates derived from the monthly transactions data as published by the Treasury. This discrepancy is illustrated in Figure 1, which shows the total foreign holdings of U.S. agency bonds. The round dots show the reported survey holdings for seven surveys from 1994 to 2006. The lines starting from each dot show the cumulated monthly transactions starting from the survey value and continuing up until the subsequent survey. As the figure shows, in every case the cumulation of the monthly positions noticeably overstates the position at the time of the next survey.

There are several reasons for the discrepancy between the reported survey holdings and cumulated monthly positions from the net transactions data. One is that the underlying monthly transactions data cannot account for all changes in holdings of securities included in the periodic surveys. For example, the transactions data for agency bonds do not include repayment flows of principal on asset-backed agency securities. We are able to adjust the transactions data starting in 2002 for these and other discrepancies, including principal repayment flows on asset-backed corporate securities, acquisitions of equity through stock swaps, and transactions in nonmarketable treasury bonds; the results using adjusted net flows for agency bonds are shown in Figure 1a. (See Appendix A for details on the adjustments for stock swaps and asset-backed securities.)

Figure 1. Total foreign holdings of U.S. agency bonds
Estimated monthly positions using unadjusted net flows in billions of dollars

Data for Figure 1 immediately follows.

Values constructed from periodic surveys of holdings and unadjusted monthly net transactions.

Data for Figure 1

Date	Survey value	Survey date: 1994-Dec	Survey date: 2000-Mar	Survey date: 2002-Jun	Survey date: 2003-Jun	Survey date: 2004-Jun	Survey date: 2005-Jun
1994-Dec	107	107	-	-	-	-	-
1995-Jan	-	109	-	-	-	-	-
1995-Feb	-	111	-	-	-	-	-
1995-Mar	-	114	-	-	-	-	-
1995-Apr	-	116	-	-	-	-	-
1995-May	-	117	-	-	-	-	-
1995-Jun	-	120	-	-	-	-	-
1995-Jul	-	123	-	-	-	-	-
1995-Aug	-	127	-	-	-	-	-
1995-Sep	-	132	-	-	-	-	-
1995-Oct	-	134	-	-	-	-	-
1995-Nov	-	138	-	-	-	-	-
1995-Dec	-	136	-	-	-	-	-
1996-Jan	-	139	-	-	-	-	-
1996-Feb	-	142	-	-	-	-	-
1996-Mar	-	142	-	-	-	-	-
1996-Apr	-	144	-	-	-	-	-
1996-May	-	147	-	-	-	-	-
1996-Jun	-	151	-	-	-	-	-
1996-Jul	-	155	-	-	-	-	-
1996-Aug	-	160	-	-	-	-	-
1996-Sep	-	164	-	-	-	-	-
1996-Oct	-	168	-	-	-	-	-
1996-Nov	-	175	-	-	-	-	-
1996-Dec	-	178	-	-	-	-	-
1997-Jan	-	182	-	-	-	-	-
1997-Feb	-	187	-	-	-	-	-
1997-Mar	-	191	-	-	-	-	-
1997-Apr	-	197	-	-	-	-	-
1997-May	-	200	-	-	-	-	-
1997-Jun	-	202	-	-	-	-	-
1997-Jul	-	208	-	-	-	-	-
1997-Aug	-	216	-	-	-	-	-
1997-Sep	-	217	-	-	-	-	-
1997-Oct	-	225	-	-	-	-	-
1997-Nov	-	225	-	-	-	-	-
1997-Dec	-	227	-	-	-	-	-
1998-Jan	-	231	-	-	-	-	-
1998-Feb	-	240	-	-	-	-	-
1998-Mar	-	249	-	-	-	-	-
1998-Apr	-	257	-	-	-	-	-
1998-May	-	260	-	-	-	-	-
1998-Jun	-	267	-	-	-	-	-
1998-Jul	-	269	-	-	-	-	-
1998-Aug	-	273	-	-	-	-	-
1998-Sep	-	273	-	-	-	-	-
1998-Oct	-	267	-	-	-	-	-
1998-Nov	-	277	-	-	-	-	-
1998-Dec	-	284	-	-	-	-	-
1999-Jan	-	293	-	-	-	-	-
1999-Feb	-	297	-	-	-	-	-
1999-Mar	-	308	-	-	-	-	-
1999-Apr	-	320	-	-	-	-	-
1999-May	-	326	-	-	-	-	-
1999-Jun	-	331	-	-	-	-	-
1999-Jul	-	339	-	-	-	-	-
1999-Aug	-	344	-	-	-	-	-
1999-Sep	-	354	-	-	-	-	-
1999-Oct	-	363	-	-	-	-	-
1999-Nov	-	371	-	-	-	-	-
1999-Dec	-	376	-	-	-	-	-
2000-Jan	-	382	-	-	-	-	-
2000-Feb	-	396	-	-	-	-	-
2000-Mar	261	411	261	-	-	-	-
2000-Apr	-	-	270	-	-	-	-
2000-May	-	-	284	-	-	-	-
2000-Jun	-	-	291	-	-	-	-
2000-Jul	-	-	301	-	-	-	-
2000-Aug	-	-	317	-	-	-	-
2000-Sep	-	-	333	-	-	-	-
2000-Oct	-	-	352	-	-	-	-
2000-Nov	-	-	370	-	-	-	-
2000-Dec	-	-	380	-	-	-	-
2001-Jan	-	-	393	-	-	-	-
2001-Feb	-	-	403	-	-	-	-
2001-Mar	-	-	422	-	-	-	-
2001-Apr	-	-	438	-	-	-	-
2001-May	-	-	445	-	-	-	-
2001-Jun	-	-	462	-	-	-	-
2001-Jul	-	-	474	-	-	-	-
2001-Aug	-	-	486	-	-	-	-
2001-Sep	-	-	494	-	-	-	-
2001-Oct	-	-	521	-	-	-	-
2001-Nov	-	-	535	-	-	-	-
2001-Dec	-	-	544	-	-	-	-
2002-Jan	-	-	552	-	-	-	-
2002-Feb	-	-	558	-	-	-	-
2002-Mar	-	-	578	-	-	-	-
2002-Apr	-	-	602	-	-	-	-
2002-May	-	-	620	-	-	-	-
2002-Jun	492	-	632	492	-	-	-
2002-Jul	-	-	-	504	-	-	-
2002-Aug	-	-	-	522	-	-	-
2002-Sep	-	-	-	543	-	-	-
2002-Oct	-	-	-	565	-	-	-
2002-Nov	-	-	-	583	-	-	-
2002-Dec	-	-	-	599	-	-	-
2003-Jan	-	-	-	623	-	-	-
2003-Feb	-	-	-	632	-	-	-
2003-Mar	-	-	-	647	-	-	-
2003-Apr	-	-	-	665	-	-	-
2003-May	-	-	-	691	-	-	-
2003-Jun	586	-	-	698	586	-	-
2003-Jul	-	-	-	-	598	-	-
2003-Aug	-	-	-	-	607	-	-
2003-Sep	-	-	-	-	607	-	-
2003-Oct	-	-	-	-	618	-	-
2003-Nov	-	-	-	-	627	-	-
2003-Dec	-	-	-	-	643	-	-
2004-Jan	-	-	-	-	669	-	-
2004-Feb	-	-	-	-	689	-	-
2004-Mar	-	-	-	-	688	-	-
2004-Apr	-	-	-	-	714	-	-
2004-May	-	-	-	-	734	-	-
2004-Jun	619	-	-	-	747	619	-
2004-Jul	-	-	-	-	-	639	-
2004-Aug	-	-	-	-	-	656	-
2004-Sep	-	-	-	-	-	665	-
2004-Oct	-	-	-	-	-	687	-
2004-Nov	-	-	-	-	-	714	-
2004-Dec	-	-	-	-	-	741	-
2005-Jan	-	-	-	-	-	766	-
2005-Feb	-	-	-	-	-	781	-
2005-Mar	-	-	-	-	-	788	-
2005-Apr	-	-	-	-	-	795	-
2005-May	-	-	-	-	-	817	-
2005-Jun	791	-	-	-	-	835	791
2005-Jul	-	-	-	-	-	-	825
2005-Aug	-	-	-	-	-	-	841
2005-Sep	-	-	-	-	-	-	860
2005-Oct	-	-	-	-	-	-	893
2005-Nov	-	-	-	-	-	-	904
2005-Dec	-	-	-	-	-	-	915
2006-Jan	-	-	-	-	-	-	945
2006-Feb	-	-	-	-	-	-	975
2006-Mar	-	-	-	-	-	-	990
2006-Apr	-	-	-	-	-	-	1008
2006-May	-	-	-	-	-	-	1044
2006-Jun	984	-	-	-	-	-	1067

Figure 1a. Total foreign holdings of U.S. agency bonds
Estimated monthly positions using adjusted net flows in billions of dollars

Data for Figure 1a immediately follows.

Values constructed from periodic surveys of holdings and adjusted monthly net transactions.

Data for Figure 1a

Date	Survey value	Survey date 1994-Dec: unadj. net flows	Survey date 2000-Mar: unadj. net flows	Survey date 2002-Jun: unadj. net flows	Survey date 2003-Jun: unadj. net flows	Survey date 2004-Jun: unadj. net flows	Survey date 2005-Jun: unadj. net flows	Survey date 1994-Dec: adj. net flows	Survey date 2000-Mar: adj. net flow	Survey date 2002-Jun: adj. net flows	Survey date 2003-Jun: adj. net flows	Survey date 2004-Jun: adj. net flows	Survey date 2005-Jun: adj. net flows
1994-Dec	107	107	-	-	-	-	-	107	-	-	-	-	-
1995-Jan	-	109	-	-	-	-	-	109	-	-	-	-	-
1995-Feb	-	111	-	-	-	-	-	111	-	-	-	-	-
1995-Mar	-	114	-	-	-	-	-	114	-	-	-	-	-
1995-Apr	-	116	-	-	-	-	-	116	-	-	-	-	-
1995-May	-	117	-	-	-	-	-	117	-	-	-	-	-
1995-Jun	-	120	-	-	-	-	-	120	-	-	-	-	-
1995-Jul	-	123	-	-	-	-	-	123	-	-	-	-	-
1995-Aug	-	127	-	-	-	-	-	127	-	-	-	-	-
1995-Sep	-	132	-	-	-	-	-	132	-	-	-	-	-
1995-Oct	-	134	-	-	-	-	-	134	-	-	-	-	-
1995-Nov	-	138	-	-	-	-	-	138	-	-	-	-	-
1995-Dec	-	136	-	-	-	-	-	136	-	-	-	-	-
1996-Jan	-	139	-	-	-	-	-	139	-	-	-	-	-
1996-Feb	-	142	-	-	-	-	-	142	-	-	-	-	-
1996-Mar	-	142	-	-	-	-	-	142	-	-	-	-	-
1996-Apr	-	144	-	-	-	-	-	144	-	-	-	-	-
1996-May	-	147	-	-	-	-	-	147	-	-	-	-	-
1996-Jun	-	151	-	-	-	-	-	151	-	-	-	-	-
1996-Jul	-	155	-	-	-	-	-	155	-	-	-	-	-
1996-Aug	-	160	-	-	-	-	-	160	-	-	-	-	-
1996-Sep	-	164	-	-	-	-	-	164	-	-	-	-	-
1996-Oct	-	168	-	-	-	-	-	168	-	-	-	-	-
1996-Nov	-	175	-	-	-	-	-	175	-	-	-	-	-
1996-Dec	-	178	-	-	-	-	-	178	-	-	-	-	-
1997-Jan	-	182	-	-	-	-	-	182	-	-	-	-	-
1997-Feb	-	187	-	-	-	-	-	187	-	-	-	-	-
1997-Mar	-	191	-	-	-	-	-	191	-	-	-	-	-
1997-Apr	-	197	-	-	-	-	-	197	-	-	-	-	-
1997-May	-	200	-	-	-	-	-	200	-	-	-	-	-
1997-Jun	-	202	-	-	-	-	-	202	-	-	-	-	-
1997-Jul	-	208	-	-	-	-	-	208	-	-	-	-	-
1997-Aug	-	216	-	-	-	-	-	216	-	-	-	-	-
1997-Sep	-	217	-	-	-	-	-	217	-	-	-	-	-
1997-Oct	-	225	-	-	-	-	-	225	-	-	-	-	-
1997-Nov	-	225	-	-	-	-	-	225	-	-	-	-	-
1997-Dec	-	227	-	-	-	-	-	227	-	-	-	-	-
1998-Jan	-	231	-	-	-	-	-	231	-	-	-	-	-
1998-Feb	-	240	-	-	-	-	-	240	-	-	-	-	-
1998-Mar	-	249	-	-	-	-	-	249	-	-	-	-	-
1998-Apr	-	257	-	-	-	-	-	257	-	-	-	-	-
1998-May	-	260	-	-	-	-	-	260	-	-	-	-	-
1998-Jun	-	267	-	-	-	-	-	267	-	-	-	-	-
1998-Jul	-	269	-	-	-	-	-	269	-	-	-	-	-
1998-Aug	-	273	-	-	-	-	-	273	-	-	-	-	-
1998-Sep	-	273	-	-	-	-	-	273	-	-	-	-	-
1998-Oct	-	267	-	-	-	-	-	267	-	-	-	-	-
1998-Nov	-	277	-	-	-	-	-	277	-	-	-	-	-
1998-Dec	-	284	-	-	-	-	-	284	-	-	-	-	-
1999-Jan	-	293	-	-	-	-	-	293	-	-	-	-	-
1999-Feb	-	297	-	-	-	-	-	297	-	-	-	-	-
1999-Mar	-	308	-	-	-	-	-	308	-	-	-	-	-
1999-Apr	-	320	-	-	-	-	-	320	-	-	-	-	-
1999-May	-	326	-	-	-	-	-	326	-	-	-	-	-
1999-Jun	-	331	-	-	-	-	-	331	-	-	-	-	-
1999-Jul	-	339	-	-	-	-	-	339	-	-	-	-	-
1999-Aug	-	344	-	-	-	-	-	344	-	-	-	-	-
1999-Sep	-	354	-	-	-	-	-	354	-	-	-	-	-
1999-Oct	-	363	-	-	-	-	-	363	-	-	-	-	-
1999-Nov	-	371	-	-	-	-	-	371	-	-	-	-	-
1999-Dec	-	376	-	-	-	-	-	376	-	-	-	-	-
2000-Jan	-	382	-	-	-	-	-	382	-	-	-	-	-
2000-Feb	-	396	-	-	-	-	-	396	-	-	-	-	-
2000-Mar	261	411	261	-	-	-	-	411	261	-	-	-	-
2000-Apr	-	-	270	-	-	-	-	-	270	-	-	-	-
2000-May	-	-	284	-	-	-	-	-	284	-	-	-	-
2000-Jun	-	-	291	-	-	-	-	-	291	-	-	-	-
2000-Jul	-	-	301	-	-	-	-	-	301	-	-	-	-
2000-Aug	-	-	317	-	-	-	-	-	317	-	-	-	-
2000-Sep	-	-	333	-	-	-	-	-	333	-	-	-	-
2000-Oct	-	-	352	-	-	-	-	-	352	-	-	-	-
2000-Nov	-	-	370	-	-	-	-	-	370	-	-	-	-
2000-Dec	-	-	380	-	-	-	-	-	380	-	-	-	-
2001-Jan	-	-	393	-	-	-	-	-	393	-	-	-	-
2001-Feb	-	-	403	-	-	-	-	-	403	-	-	-	-
2001-Mar	-	-	422	-	-	-	-	-	422	-	-	-	-
2001-Apr	-	-	438	-	-	-	-	-	438	-	-	-	-
2001-May	-	-	445	-	-	-	-	-	445	-	-	-	-
2001-Jun	-	-	462	-	-	-	-	-	462	-	-	-	-
2001-Jul	-	-	474	-	-	-	-	-	474	-	-	-	-
2001-Aug	-	-	486	-	-	-	-	-	486	-	-	-	-
2001-Sep	-	-	494	-	-	-	-	-	494	-	-	-	-
2001-Oct	-	-	521	-	-	-	-	-	521	-	-	-	-
2001-Nov	-	-	535	-	-	-	-	-	535	-	-	-	-
2001-Dec	-	-	544	-	-	-	-	-	544	-	-	-	-
2002-Jan	-	-	552	-	-	-	-	-	552	-	-	-	-
2002-Feb	-	-	558	-	-	-	-	-	558	-	-	-	-
2002-Mar	-	-	578	-	-	-	-	-	578	-	-	-	-
2002-Apr	-	-	602	-	-	-	-	-	602	-	-	-	-
2002-May	-	-	620	-	-	-	-	-	620	-	-	-	-
2002-Jun	492	-	632	492	-	-	-	-	632	492	-	-	-
2002-Jul	-	-	-	504	-	-	-	-	-	502	-	-	-
2002-Aug	-	-	-	522	-	-	-	-	-	516	-	-	-
2002-Sep	-	-	-	543	-	-	-	-	-	533	-	-	-
2002-Oct	-	-	-	565	-	-	-	-	-	550	-	-	-
2002-Nov	-	-	-	583	-	-	-	-	-	562	-	-	-
2002-Dec	-	-	-	599	-	-	-	-	-	572	-	-	-
2003-Jan	-	-	-	623	-	-	-	-	-	590	-	-	-
2003-Feb	-	-	-	632	-	-	-	-	-	593	-	-	-
2003-Mar	-	-	-	647	-	-	-	-	-	602	-	-	-
2003-Apr	-	-	-	665	-	-	-	-	-	614	-	-	-
2003-May	-	-	-	691	-	-	-	-	-	632	-	-	-
2003-Jun	586	-	-	698	586	-	-	-	-	633	586	-	-
2003-Jul	-	-	-	-	598	-	-	-	-	-	590	-	-
2003-Aug	-	-	-	-	607	-	-	-	-	-	589	-	-
2003-Sep	-	-	-	-	607	-	-	-	-	-	581	-	-
2003-Oct	-	-	-	-	618	-	-	-	-	-	587	-	-
2003-Nov	-	-	-	-	627	-	-	-	-	-	593	-	-
2003-Dec	-	-	-	-	643	-	-	-	-	-	604	-	-
2004-Jan	-	-	-	-	669	-	-	-	-	-	626	-	-
2004-Feb	-	-	-	-	689	-	-	-	-	-	644	-	-
2004-Mar	-	-	-	-	688	-	-	-	-	-	639	-	-
2004-Apr	-	-	-	-	714	-	-	-	-	-	661	-	-
2004-May	-	-	-	-	734	-	-	-	-	-	674	-	-
2004-Jun	619	-	-	-	747	619	-	-	-	-	684	619	-
2004-Jul	-	-	-	-	-	639	-	-	-	-	-	635	-
2004-Aug	-	-	-	-	-	656	-	-	-	-	-	649	-
2004-Sep	-	-	-	-	-	665	-	-	-	-	-	653	-
2004-Oct	-	-	-	-	-	687	-	-	-	-	-	671	-
2004-Nov	-	-	-	-	-	714	-	-	-	-	-	695	-
2004-Dec	-	-	-	-	-	741	-	-	-	-	-	717	-
2005-Jan	-	-	-	-	-	766	-	-	-	-	-	738	-
2005-Feb	-	-	-	-	-	781	-	-	-	-	-	749	-
2005-Mar	-	-	-	-	-	788	-	-	-	-	-	752	-
2005-Apr	-	-	-	-	-	795	-	-	-	-	-	754	-
2005-May	-	-	-	-	-	817	-	-	-	-	-	770	-
2005-Jun	791	-	-	-	-	835	791	-	-	-	-	784	791
2005-Jul	-	-	-	-	-	-	825	-	-	-	-	-	819
2005-Aug	-	-	-	-	-	-	841	-	-	-	-	-	829
2005-Sep	-	-	-	-	-	-	860	-	-	-	-	-	841
2005-Oct	-	-	-	-	-	-	893	-	-	-	-	-	868
2005-Nov	-	-	-	-	-	-	904	-	-	-	-	-	873
2005-Dec	-	-	-	-	-	-	915	-	-	-	-	-	879
2006-Jan	-	-	-	-	-	-	945	-	-	-	-	-	904
2006-Feb	-	-	-	-	-	-	975	-	-	-	-	-	930
2006-Mar	-	-	-	-	-	-	990	-	-	-	-	-	941
2006-Apr	-	-	-	-	-	-	1008	-	-	-	-	-	954
2006-May	-	-	-	-	-	-	1044	-	-	-	-	-	985
2006-Jun	984	-	-	-	-	-	1067	-	-	-	-	-	1002

As expected, using the adjusted net flows bring the estimated positions closer to the survey results, but a noticeable gap still remains. Another factor that contributes to the discrepancy is that the transactions data do not reflect any valuation changes, which do affect the measured positions reported in the periodic surveys. Figure 2 reports total foreign holdings of treasury bonds, agency bonds, corporate bonds, and equities estimated using adjusted net flows with and without allowance for valuation changes. (See Appendix B for details on how the allowance for valuation changes was made.)

Figure 2. Total foreign holdings of U.S. long-term securities
Estimated monthly positions using adjusted net flows in billions of dollars

Data for Figure 2 immediately follows.

Values constructed from periodic surveys of holdings, adjusted monthly net transactions and valuation adjustments.

Data for Figure 2 - Treasury Bonds

Date	Survey value	Survey date 2000-Mar: adj. net flows	Survey date 2002-Jun: adj. net flows	Survey date 2003-Jun: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2000-Mar: with valuation adj.	Survey date 2002-Jun: with valuation adj.	Survey date 2003-Jun: with valuation adj.	Survey date 2004-Jun: with valuation adj.	Survey date 2005-Jun: with valuation adj.
2000-Mar	884	884	-	-	-	-	884	-	-	-	-
2000-Apr	-	899	-	-	-	-	891	-	-	-	-
2000-May	-	892	-	-	-	-	881	-	-	-	-
2000-Jun	-	874	-	-	-	-	873	-	-	-	-
2000-Jul	-	868	-	-	-	-	871	-	-	-	-
2000-Aug	-	868	-	-	-	-	879	-	-	-	-
2000-Sep	-	859	-	-	-	-	866	-	-	-	-
2000-Oct	-	856	-	-	-	-	867	-	-	-	-
2000-Nov	-	842	-	-	-	-	866	-	-	-	-
2000-Dec	-	832	-	-	-	-	869	-	-	-	-
2001-Jan	-	823	-	-	-	-	862	-	-	-	-
2001-Feb	-	831	-	-	-	-	876	-	-	-	-
2001-Mar	-	836	-	-	-	-	879	-	-	-	-
2001-Apr	-	822	-	-	-	-	850	-	-	-	-
2001-May	-	825	-	-	-	-	852	-	-	-	-
2001-Jun	-	821	-	-	-	-	848	-	-	-	-
2001-Jul	-	810	-	-	-	-	854	-	-	-	-
2001-Aug	-	815	-	-	-	-	866	-	-	-	-
2001-Sep	-	814	-	-	-	-	874	-	-	-	-
2001-Oct	-	829	-	-	-	-	910	-	-	-	-
2001-Nov	-	842	-	-	-	-	895	-	-	-	-
2001-Dec	-	851	-	-	-	-	891	-	-	-	-
2002-Jan	-	835	-	-	-	-	878	-	-	-	-
2002-Feb	-	844	-	-	-	-	891	-	-	-	-
2002-Mar	-	863	-	-	-	-	884	-	-	-	-
2002-Apr	-	855	-	-	-	-	893	-	-	-	-
2002-May	-	865	-	-	-	-	904	-	-	-	-
2002-Jun	905	879	905	-	-	-	927	905	-	-	-
2002-Jul	-	-	929	-	-	-	-	947	-	-	-
2002-Aug	-	-	934	-	-	-	-	967	-	-	-
2002-Sep	-	-	955	-	-	-	-	1011	-	-	-
2002-Oct	-	-	962	-	-	-	-	1002	-	-	-
2002-Nov	-	-	983	-	-	-	-	1009	-	-	-
2002-Dec	-	-	996	-	-	-	-	1044	-	-	-
2003-Jan	-	-	998	-	-	-	-	1038	-	-	-
2003-Feb	-	-	995	-	-	-	-	1049	-	-	-
2003-Mar	-	-	1022	-	-	-	-	1067	-	-	-
2003-Apr	-	-	1030	-	-	-	-	1077	-	-	-
2003-May	-	-	1072	-	-	-	-	1145	-	-	-
2003-Jun	1114	-	1112	1114	-	-	-	1173	1114	-	-
2003-Jul	-	-	-	1164	-	-	-	-	1111	-	-
2003-Aug	-	-	-	1181	-	-	-	-	1130	-	-
2003-Sep	-	-	-	1188	-	-	-	-	1167	-	-
2003-Oct	-	-	-	1198	-	-	-	-	1155	-	-
2003-Nov	-	-	-	1234	-	-	-	-	1187	-	-
2003-Dec	-	-	-	1262	-	-	-	-	1221	-	-
2004-Jan	-	-	-	1308	-	-	-	-	1273	-	-
2004-Feb	-	-	-	1334	-	-	-	-	1311	-	-
2004-Mar	-	-	-	1385	-	-	-	-	1369	-	-
2004-Apr	-	-	-	1435	-	-	-	-	1371	-	-
2004-May	-	-	-	1464	-	-	-	-	1390	-	-
2004-Jun	1424	-	-	1509	1424	-	-	-	1435	1424	-
2004-Jul	-	-	-	-	1437	-	-	-	-	1445	-
2004-Aug	-	-	-	-	1451	-	-	-	-	1483	-
2004-Sep	-	-	-	-	1468	-	-	-	-	1499	-
2004-Oct	-	-	-	-	1487	-	-	-	-	1525	-
2004-Nov	-	-	-	-	1520	-	-	-	-	1532	-
2004-Dec	-	-	-	-	1529	-	-	-	-	1550	-
2005-Jan	-	-	-	-	1562	-	-	-	-	1589	-
2005-Feb	-	-	-	-	1606	-	-	-	-	1615	-
2005-Mar	-	-	-	-	1641	-	-	-	-	1638	-
2005-Apr	-	-	-	-	1670	-	-	-	-	1691	-
2005-May	-	-	-	-	1672	-	-	-	-	1707	-
2005-Jun	1598	-	-	-	1692	1598	-	-	-	1731	1598
2005-Jul	-	-	-	-	-	1626	-	-	-	-	1599
2005-Aug	-	-	-	-	-	1652	-	-	-	-	1644
2005-Sep	-	-	-	-	-	1674	-	-	-	-	1639
2005-Oct	-	-	-	-	-	1704	-	-	-	-	1650
2005-Nov	-	-	-	-	-	1756	-	-	-	-	1704
2005-Dec	-	-	-	-	-	1773	-	-	-	-	1732
2006-Jan	-	-	-	-	-	1775	-	-	-	-	1722
2006-Feb	-	-	-	-	-	1794	-	-	-	-	1737
2006-Mar	-	-	-	-	-	1803	-	-	-	-	1721
2006-Apr	-	-	-	-	-	1799	-	-	-	-	1703
2006-May	-	-	-	-	-	1815	-	-	-	-	1712
2006-Jun	1726	-	-	-	-	1844	-	-	-	-	1740

Data for Figure 2 - Agency Bonds

Date	Survey value	Survey date 2000-Mar: adj. net flows	Survey date 2002-Jun: adj. net flows	Survey date 2003-Jun: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2000-Mar: with valuation adj.	Survey date 2002-Jun: with valuation adj.	Survey date 2003-Jun: with valuation adj.	Survey date 2004-Jun: with valuation adj.	Survey date 2005-Jun: with valuation adj.
2000-Mar	261	261	-	-	-	-	261	-	-	-	-
2000-Apr	-	270	-	-	-	-	269	-	-	-	-
2000-May	-	284	-	-	-	-	280	-	-	-	-
2000-Jun	-	291	-	-	-	-	291	-	-	-	-
2000-Jul	-	301	-	-	-	-	302	-	-	-	-
2000-Aug	-	317	-	-	-	-	321	-	-	-	-
2000-Sep	-	333	-	-	-	-	339	-	-	-	-
2000-Oct	-	352	-	-	-	-	358	-	-	-	-
2000-Nov	-	370	-	-	-	-	381	-	-	-	-
2000-Dec	-	380	-	-	-	-	398	-	-	-	-
2001-Jan	-	393	-	-	-	-	415	-	-	-	-
2001-Feb	-	403	-	-	-	-	426	-	-	-	-
2001-Mar	-	422	-	-	-	-	446	-	-	-	-
2001-Apr	-	438	-	-	-	-	457	-	-	-	-
2001-May	-	445	-	-	-	-	463	-	-	-	-
2001-Jun	-	462	-	-	-	-	479	-	-	-	-
2001-Jul	-	474	-	-	-	-	500	-	-	-	-
2001-Aug	-	486	-	-	-	-	515	-	-	-	-
2001-Sep	-	494	-	-	-	-	531	-	-	-	-
2001-Oct	-	521	-	-	-	-	567	-	-	-	-
2001-Nov	-	535	-	-	-	-	569	-	-	-	-
2001-Dec	-	544	-	-	-	-	571	-	-	-	-
2002-Jan	-	552	-	-	-	-	580	-	-	-	-
2002-Feb	-	558	-	-	-	-	589	-	-	-	-
2002-Mar	-	578	-	-	-	-	596	-	-	-	-
2002-Apr	-	602	-	-	-	-	631	-	-	-	-
2002-May	-	620	-	-	-	-	651	-	-	-	-
2002-Jun	492	632	492	-	-	-	668	492	-	-	-
2002-Jul	-	-	502	-	-	-	-	508	-	-	-
2002-Aug	-	-	516	-	-	-	-	527	-	-	-
2002-Sep	-	-	533	-	-	-	-	550	-	-	-
2002-Oct	-	-	550	-	-	-	-	564	-	-	-
2002-Nov	-	-	562	-	-	-	-	571	-	-	-
2002-Dec	-	-	572	-	-	-	-	589	-	-	-
2003-Jan	-	-	590	-	-	-	-	605	-	-	-
2003-Feb	-	-	593	-	-	-	-	614	-	-	-
2003-Mar	-	-	602	-	-	-	-	621	-	-	-
2003-Apr	-	-	614	-	-	-	-	633	-	-	-
2003-May	-	-	632	-	-	-	-	659	-	-	-
2003-Jun	586	-	633	586	-	-	-	656	586	-	-
2003-Jul	-	-	-	590	-	-	-	-	570	-	-
2003-Aug	-	-	-	589	-	-	-	-	570	-	-
2003-Sep	-	-	-	581	-	-	-	-	574	-	-
2003-Oct	-	-	-	587	-	-	-	-	573	-	-
2003-Nov	-	-	-	593	-	-	-	-	577	-	-
2003-Dec	-	-	-	604	-	-	-	-	591	-	-
2004-Jan	-	-	-	626	-	-	-	-	616	-	-
2004-Feb	-	-	-	644	-	-	-	-	638	-	-
2004-Mar	-	-	-	639	-	-	-	-	635	-	-
2004-Apr	-	-	-	661	-	-	-	-	640	-	-
2004-May	-	-	-	674	-	-	-	-	649	-	-
2004-Jun	619	-	-	684	619	-	-	-	659	619	-
2004-Jul	-	-	-	-	635	-	-	-	-	638	-
2004-Aug	-	-	-	-	649	-	-	-	-	660	-
2004-Sep	-	-	-	-	653	-	-	-	-	663	-
2004-Oct	-	-	-	-	671	-	-	-	-	684	-
2004-Nov	-	-	-	-	695	-	-	-	-	701	-
2004-Dec	-	-	-	-	717	-	-	-	-	725	-
2005-Jan	-	-	-	-	738	-	-	-	-	746	-
2005-Feb	-	-	-	-	749	-	-	-	-	751	-
2005-Mar	-	-	-	-	752	-	-	-	-	749	-
2005-Apr	-	-	-	-	754	-	-	-	-	757	-
2005-May	-	-	-	-	770	-	-	-	-	778	-
2005-Jun	791	-	-	-	784	791	-	-	-	792	791
2005-Jul	-	-	-	-	-	819	-	-	-	-	811
2005-Aug	-	-	-	-	-	829	-	-	-	-	826
2005-Sep	-	-	-	-	-	841	-	-	-	-	829
2005-Oct	-	-	-	-	-	868	-	-	-	-	849
2005-Nov	-	-	-	-	-	873	-	-	-	-	854
2005-Dec	-	-	-	-	-	879	-	-	-	-	863
2006-Jan	-	-	-	-	-	904	-	-	-	-	886
2006-Feb	-	-	-	-	-	930	-	-	-	-	911
2006-Mar	-	-	-	-	-	941	-	-	-	-	913
2006-Apr	-	-	-	-	-	954	-	-	-	-	922
2006-May	-	-	-	-	-	985	-	-	-	-	949
2006-Jun	984	-	-	-	-	1002	-	-	-	-	964

Data for Figure 2 - Corporate Bonds

Date	Survey value	Survey date 2000-Mar: adj. net flows	Survey date 2002-Jun: adj. net flows	Survey date 2003-Jun: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2000-Mar: with valuation adj.	Survey date 2002-Jun: with valuation adj.	Survey date 2003-Jun: with valuation adj.	Survey date 2004-Jun: with valuation adj.	Survey date 2005-Jun: with valuation adj.
2000-Mar	703	703	-	-	-	-	703	-	-	-	-
2000-Apr	-	712	-	-	-	-	703	-	-	-	-
2000-May	-	720	-	-	-	-	704	-	-	-	-
2000-Jun	-	746	-	-	-	-	740	-	-	-	-
2000-Jul	-	756	-	-	-	-	754	-	-	-	-
2000-Aug	-	778	-	-	-	-	779	-	-	-	-
2000-Sep	-	792	-	-	-	-	794	-	-	-	-
2000-Oct	-	805	-	-	-	-	803	-	-	-	-
2000-Nov	-	824	-	-	-	-	828	-	-	-	-
2000-Dec	-	843	-	-	-	-	857	-	-	-	-
2001-Jan	-	857	-	-	-	-	887	-	-	-	-
2001-Feb	-	886	-	-	-	-	918	-	-	-	-
2001-Mar	-	913	-	-	-	-	946	-	-	-	-
2001-Apr	-	935	-	-	-	-	959	-	-	-	-
2001-May	-	968	-	-	-	-	995	-	-	-	-
2001-Jun	-	983	-	-	-	-	1009	-	-	-	-
2001-Jul	-	998	-	-	-	-	1044	-	-	-	-
2001-Aug	-	1011	-	-	-	-	1064	-	-	-	-
2001-Sep	-	1021	-	-	-	-	1064	-	-	-	-
2001-Oct	-	1035	-	-	-	-	1097	-	-	-	-
2001-Nov	-	1051	-	-	-	-	1102	-	-	-	-
2001-Dec	-	1065	-	-	-	-	1101	-	-	-	-
2002-Jan	-	1080	-	-	-	-	1119	-	-	-	-
2002-Feb	-	1088	-	-	-	-	1127	-	-	-	-
2002-Mar	-	1115	-	-	-	-	1130	-	-	-	-
2002-Apr	-	1132	-	-	-	-	1155	-	-	-	-
2002-May	-	1159	-	-	-	-	1188	-	-	-	-
2002-Jun	1130	1180	1130	-	-	-	1207	1130	-	-	-
2002-Jul	-	-	1132	-	-	-	-	1129	-	-	-
2002-Aug	-	-	1145	-	-	-	-	1162	-	-	-
2002-Sep	-	-	1146	-	-	-	-	1179	-	-	-
2002-Oct	-	-	1154	-	-	-	-	1169	-	-	-
2002-Nov	-	-	1173	-	-	-	-	1193	-	-	-
2002-Dec	-	-	1183	-	-	-	-	1230	-	-	-
2003-Jan	-	-	1203	-	-	-	-	1250	-	-	-
2003-Feb	-	-	1213	-	-	-	-	1276	-	-	-
2003-Mar	-	-	1237	-	-	-	-	1294	-	-	-
2003-Apr	-	-	1255	-	-	-	-	1327	-	-	-
2003-May	-	-	1280	-	-	-	-	1384	-	-	-
2003-Jun	1236	-	1296	1236	-	-	-	1392	1236	-	-
2003-Jul	-	-	-	1259	-	-	-	-	1208	-	-
2003-Aug	-	-	-	1272	-	-	-	-	1223	-	-
2003-Sep	-	-	-	1288	-	-	-	-	1271	-	-
2003-Oct	-	-	-	1303	-	-	-	-	1266	-	-
2003-Nov	-	-	-	1326	-	-	-	-	1289	-	-
2003-Dec	-	-	-	1342	-	-	-	-	1313	-	-
2004-Jan	-	-	-	1350	-	-	-	-	1326	-	-
2004-Feb	-	-	-	1367	-	-	-	-	1353	-	-
2004-Mar	-	-	-	1391	-	-	-	-	1383	-	-
2004-Apr	-	-	-	1405	-	-	-	-	1355	-	-
2004-May	-	-	-	1422	-	-	-	-	1357	-	-
2004-Jun	1455	-	-	1447	1455	-	-	-	1381	1455	-
2004-Jul	-	-	-	-	1480	-	-	-	-	1489	-
2004-Aug	-	-	-	-	1502	-	-	-	-	1538	-
2004-Sep	-	-	-	-	1545	-	-	-	-	1582	-
2004-Oct	-	-	-	-	1562	-	-	-	-	1606	-
2004-Nov	-	-	-	-	1585	-	-	-	-	1612	-
2004-Dec	-	-	-	-	1623	-	-	-	-	1660	-
2005-Jan	-	-	-	-	1637	-	-	-	-	1679	-
2005-Feb	-	-	-	-	1663	-	-	-	-	1689	-
2005-Mar	-	-	-	-	1679	-	-	-	-	1683	-
2005-Apr	-	-	-	-	1691	-	-	-	-	1709	-
2005-May	-	-	-	-	1704	-	-	-	-	1735	-
2005-Jun	1729	-	-	-	1748	1729	-	-	-	1786	1729
2005-Jul	-	-	-	-	-	1744	-	-	-	-	1720
2005-Aug	-	-	-	-	-	1772	-	-	-	-	1765
2005-Sep	-	-	-	-	-	1808	-	-	-	-	1771
2005-Oct	-	-	-	-	-	1836	-	-	-	-	1774
2005-Nov	-	-	-	-	-	1867	-	-	-	-	1805
2005-Dec	-	-	-	-	-	1896	-	-	-	-	1842
2006-Jan	-	-	-	-	-	1919	-	-	-	-	1854
2006-Feb	-	-	-	-	-	1951	-	-	-	-	1883
2006-Mar	-	-	-	-	-	1993	-	-	-	-	1896
2006-Apr	-	-	-	-	-	2026	-	-	-	-	1915
2006-May	-	-	-	-	-	2062	-	-	-	-	1940
2006-Jun	2021	-	-	-	-	2094	-	-	-	-	1964

Data for Figure 2 - Equities

Date	Survey value	Survey date 2000-Mar: adj. net flows	Survey date 2002-Jun: adj. net flows	Survey date 2003-Jun: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2000-Mar: with valuation adj.	Survey date 2002-Jun: with valuation adj.	Survey date 2003-Jun: with valuation adj.	Survey date 2004-Jun: with valuation adj.	Survey date 2005-Jun: with valuation adj.
2000-Mar	1709	1709	-	-	-	-	1709	-	-	-	-
2000-Apr	-	1712	-	-	-	-	1656	-	-	-	-
2000-May	-	1726	-	-	-	-	1624	-	-	-	-
2000-Jun	-	1744	-	-	-	-	1679	-	-	-	-
2000-Jul	-	1762	-	-	-	-	1665	-	-	-	-
2000-Aug	-	1789	-	-	-	-	1778	-	-	-	-
2000-Sep	-	1798	-	-	-	-	1688	-	-	-	-
2000-Oct	-	1814	-	-	-	-	1691	-	-	-	-
2000-Nov	-	1826	-	-	-	-	1567	-	-	-	-
2000-Dec	-	1838	-	-	-	-	1576	-	-	-	-
2001-Jan	-	1862	-	-	-	-	1656	-	-	-	-
2001-Feb	-	1871	-	-	-	-	1515	-	-	-	-
2001-Mar	-	1879	-	-	-	-	1425	-	-	-	-
2001-Apr	-	1886	-	-	-	-	1542	-	-	-	-
2001-May	-	1904	-	-	-	-	1566	-	-	-	-
2001-Jun	-	1914	-	-	-	-	1540	-	-	-	-
2001-Jul	-	1926	-	-	-	-	1535	-	-	-	-
2001-Aug	-	1940	-	-	-	-	1565	-	-	-	-
2001-Sep	-	1928	-	-	-	-	1554	-	-	-	-
2001-Oct	-	1936	-	-	-	-	1348	-	-	-	-
2001-Nov	-	1949	-	-	-	-	1463	-	-	-	-
2001-Dec	-	1962	-	-	-	-	1483	-	-	-	-
2002-Jan	-	1971	-	-	-	-	1469	-	-	-	-
2002-Feb	-	1977	-	-	-	-	1476	-	-	-	-
2002-Mar	-	1986	-	-	-	-	1507	-	-	-	-
2002-Apr	-	1994	-	-	-	-	1415	-	-	-	-
2002-May	-	1994	-	-	-	-	1402	-	-	-	-
2002-Jun	1395	1998	1395	-	-	-	1298	1395	-	-	-
2002-Jul	-	-	1405	-	-	-	-	1303	-	-	-
2002-Aug	-	-	1410	-	-	-	-	1312	-	-	-
2002-Sep	-	-	1403	-	-	-	-	1156	-	-	-
2002-Oct	-	-	1407	-	-	-	-	1263	-	-	-
2002-Nov	-	-	1413	-	-	-	-	1343	-	-	-
2002-Dec	-	-	1416	-	-	-	-	1263	-	-	-
2003-Jan	-	-	1414	-	-	-	-	1228	-	-	-
2003-Feb	-	-	1412	-	-	-	-	1204	-	-	-
2003-Mar	-	-	1414	-	-	-	-	1217	-	-	-
2003-Apr	-	-	1419	-	-	-	-	1321	-	-	-
2003-May	-	-	1425	-	-	-	-	1396	-	-	-
2003-Jun	1564	-	1435	1564	-	-	-	1421	1564	-	-
2003-Jul	-	-	-	1555	-	-	-	-	1582	-	-
2003-Aug	-	-	-	1567	-	-	-	-	1621	-	-
2003-Sep	-	-	-	1560	-	-	-	-	1593	-	-
2003-Oct	-	-	-	1560	-	-	-	-	1682	-	-
2003-Nov	-	-	-	1569	-	-	-	-	1704	-	-
2003-Dec	-	-	-	1583	-	-	-	-	1804	-	-
2004-Jan	-	-	-	1594	-	-	-	-	1845	-	-
2004-Feb	-	-	-	1599	-	-	-	-	1870	-	-
2004-Mar	-	-	-	1589	-	-	-	-	1828	-	-
2004-Apr	-	-	-	1600	-	-	-	-	1809	-	-
2004-May	-	-	-	1595	-	-	-	-	1824	-	-
2004-Jun	1930	-	-	1599	1930	-	-	-	1860	1930	-
2004-Jul	-	-	-	-	1938	-	-	-	-	1870	-
2004-Aug	-	-	-	-	1936	-	-	-	-	1874	-
2004-Sep	-	-	-	-	1932	-	-	-	-	1888	-
2004-Oct	-	-	-	-	1935	-	-	-	-	1918	-
2004-Nov	-	-	-	-	1972	-	-	-	-	2029	-
2004-Dec	-	-	-	-	1979	-	-	-	-	2105	-
2005-Jan	-	-	-	-	1992	-	-	-	-	2063	-
2005-Feb	-	-	-	-	1997	-	-	-	-	2107	-
2005-Mar	-	-	-	-	2000	-	-	-	-	2073	-
2005-Apr	-	-	-	-	2004	-	-	-	-	2037	-
2005-May	-	-	-	-	2004	-	-	-	-	2101	-
2005-Jun	2144	-	-	-	2008	2144	-	-	-	2106	2144
2005-Jul	-	-	-	-	-	2152	-	-	-	-	2230
2005-Aug	-	-	-	-	-	2154	-	-	-	-	2206
2005-Sep	-	-	-	-	-	2177	-	-	-	-	2246
2005-Oct	-	-	-	-	-	2186	-	-	-	-	2216
2005-Nov	-	-	-	-	-	2193	-	-	-	-	2307
2005-Dec	-	-	-	-	-	2203	-	-	-	-	2315
2006-Jan	-	-	-	-	-	2229	-	-	-	-	2401
2006-Feb	-	-	-	-	-	2246	-	-	-	-	2416
2006-Mar	-	-	-	-	-	2269	-	-	-	-	2467
2006-Apr	-	-	-	-	-	2282	-	-	-	-	2509
2006-May	-	-	-	-	-	2297	-	-	-	-	2444
2006-Jun	2430	-	-	-	-	2295	-	-	-	-	2442

Again, as we would expect, allowing for valuation changes brings the estimates closer to the reported surveys, but does not eliminate the gaps entirely. Unsurprisingly, the effect of including valuation changes is much larger for equities than it is for any of the bonds.

Thus far, we have considered the challenges in constructing estimates of holdings of U.S. securities by all foreigners. An additional complication arises in constructing estimated positions for individual country holdings based on the reported monthly transactions by country because of the geographic distortion caused by financial center transaction bias (see Griever, Lee, and Warnock (2001) and Warnock and Cleaver (2002)). By design, the monthly transactions data are recorded according to country of the first cross-border counter-party, and not the country of the ultimate buyer or actual seller or issuer of the security. As a result, the monthly transactions data report purchases and sales that are concentrated in major international financial centers. For example, in 2005, nearly two-thirds of reported purchases and sales of U.S. long-term securities were recorded against the United Kingdom and Caribbean financial centers. Thus, constructing estimated positions based on the country-level monthly transactions data will tend to generate estimates of holdings by residents of such financial center locations that considerably overstate actual holdings as reported in the next survey, and will tend to underestimate holdings by residents of other countries.

Figures 3 and 3a illustrate this problem for holdings of U.S. Treasury and corporate bonds by residents of the United Kingdom⁴ and the euro area. For the U.K., estimated positions, even after adjusting for omitted securities and valuation changes, are consistently much higher than the survey results, presumably representing transactions in U.S. securities made in the U.K. on behalf of third parties.

Figure 3. U.K. holdings of U.S. long-term securities
Estimated monthly positions using adjusted net flows with valuation changes in billions of dollars

Data for Figure 3 immediately follows.

Values constructed from periodic surveys of holdings, adjusted monthly net transactions and valuation adjustments. Includes Channel Islands and the Isle of Man.

Data for Figure 3 - Treasury bonds

Date	Survey value	Survey date 2000-Mar	Survey date 2002-Jun	Survey date 2003-Jun	Survey date 2004-Jun	Survey date 2005-Jun
2000-Mar	76	76	-	-	-	-
2000-Apr	-	74	-	-	-	-
2000-May	-	74	-	-	-	-
2000-Jun	-	66	-	-	-	-
2000-Jul	-	62	-	-	-	-
2000-Aug	-	66	-	-	-	-
2000-Sep	-	63	-	-	-	-
2000-Oct	-	60	-	-	-	-
2000-Nov	-	55	-	-	-	-
2000-Dec	-	51	-	-	-	-
2001-Jan	-	49	-	-	-	-
2001-Feb	-	50	-	-	-	-
2001-Mar	-	60	-	-	-	-
2001-Apr	-	54	-	-	-	-
2001-May	-	55	-	-	-	-
2001-Jun	-	52	-	-	-	-
2001-Jul	-	45	-	-	-	-
2001-Aug	-	41	-	-	-	-
2001-Sep	-	43	-	-	-	-
2001-Oct	-	52	-	-	-	-
2001-Nov	-	43	-	-	-	-
2001-Dec	-	44	-	-	-	-
2002-Jan	-	46	-	-	-	-
2002-Feb	-	54	-	-	-	-
2002-Mar	-	63	-	-	-	-
2002-Apr	-	64	-	-	-	-
2002-May	-	75	-	-	-	-
2002-Jun	48	82	48	-	-	-
2002-Jul	-	-	61	-	-	-
2002-Aug	-	-	66	-	-	-
2002-Sep	-	-	74	-	-	-
2002-Oct	-	-	78	-	-	-
2002-Nov	-	-	77	-	-	-
2002-Dec	-	-	76	-	-	-
2003-Jan	-	-	77	-	-	-
2003-Feb	-	-	73	-	-	-
2003-Mar	-	-	77	-	-	-
2003-Apr	-	-	74	-	-	-
2003-May	-	-	77	-	-	-
2003-Jun	49	-	86	49	-	-
2003-Jul	-	-	-	69	-	-
2003-Aug	-	-	-	75	-	-
2003-Sep	-	-	-	65	-	-
2003-Oct	-	-	-	69	-	-
2003-Nov	-	-	-	69	-	-
2003-Dec	-	-	-	73	-	-
2004-Jan	-	-	-	83	-	-
2004-Feb	-	-	-	86	-	-
2004-Mar	-	-	-	93	-	-
2004-Apr	-	-	-	105	-	-
2004-May	-	-	-	103	-	-
2004-Jun	47	-	-	109	47	-
2004-Jul	-	-	-	-	52	-
2004-Aug	-	-	-	-	56	-
2004-Sep	-	-	-	-	57	-
2004-Oct	-	-	-	-	61	-
2004-Nov	-	-	-	-	77	-
2004-Dec	-	-	-	-	89	-
2005-Jan	-	-	-	-	88	-
2005-Feb	-	-	-	-	97	-
2005-Mar	-	-	-	-	109	-
2005-Apr	-	-	-	-	113	-
2005-May	-	-	-	-	122	-
2005-Jun	53	-	-	-	137	53
2005-Jul	-	-	-	-	-	67
2005-Aug	-	-	-	-	-	81
2005-Sep	-	-	-	-	-	89
2005-Oct	-	-	-	-	-	93
2005-Nov	-	-	-	-	-	128
2005-Dec	-	-	-	-	-	139
2006-Jan	-	-	-	-	-	149
2006-Feb	-	-	-	-	-	154
2006-Mar	-	-	-	-	-	168
2006-Apr	-	-	-	-	-	154
2006-May	-	-	-	-	-	162
2006-Jun	50	-	-	-	-	188

Data for Figure 3 - Corporate Bonds

Date	Survey value	Survey date 2000-Mar	Survey date 2002-Jun	Survey date 2003-Jun	Survey date 2004-Jun	Survey date 2005-Jun
2000-Mar	114	114	-	-	-	-
2000-Apr	-	117	-	-	-	-
2000-May	-	119	-	-	-	-
2000-Jun	-	137	-	-	-	-
2000-Jul	-	144	-	-	-	-
2000-Aug	-	158	-	-	-	-
2000-Sep	-	167	-	-	-	-
2000-Oct	-	174	-	-	-	-
2000-Nov	-	185	-	-	-	-
2000-Dec	-	198	-	-	-	-
2001-Jan	-	210	-	-	-	-
2001-Feb	-	226	-	-	-	-
2001-Mar	-	242	-	-	-	-
2001-Apr	-	250	-	-	-	-
2001-May	-	269	-	-	-	-
2001-Jun	-	274	-	-	-	-
2001-Jul	-	286	-	-	-	-
2001-Aug	-	294	-	-	-	-
2001-Sep	-	297	-	-	-	-
2001-Oct	-	304	-	-	-	-
2001-Nov	-	310	-	-	-	-
2001-Dec	-	313	-	-	-	-
2002-Jan	-	324	-	-	-	-
2002-Feb	-	326	-	-	-	-
2002-Mar	-	333	-	-	-	-
2002-Apr	-	342	-	-	-	-
2002-May	-	360	-	-	-	-
2002-Jun	108	370	108	-	-	-
2002-Jul	-	-	108	-	-	-
2002-Aug	-	-	112	-	-	-
2002-Sep	-	-	117	-	-	-
2002-Oct	-	-	119	-	-	-
2002-Nov	-	-	130	-	-	-
2002-Dec	-	-	139	-	-	-
2003-Jan	-	-	149	-	-	-
2003-Feb	-	-	156	-	-	-
2003-Mar	-	-	171	-	-	-
2003-Apr	-	-	186	-	-	-
2003-May	-	-	205	-	-	-
2003-Jun	122	-	209	122	-	-
2003-Jul	-	-	-	126	-	-
2003-Aug	-	-	-	130	-	-
2003-Sep	-	-	-	145	-	-
2003-Oct	-	-	-	149	-	-
2003-Nov	-	-	-	162	-	-
2003-Dec	-	-	-	173	-	-
2004-Jan	-	-	-	173	-	-
2004-Feb	-	-	-	178	-	-
2004-Mar	-	-	-	187	-	-
2004-Apr	-	-	-	189	-	-
2004-May	-	-	-	192	-	-
2004-Jun	183	-	-	197	183	-
2004-Jul	-	-	-	-	194	-
2004-Aug	-	-	-	-	207	-
2004-Sep	-	-	-	-	227	-
2004-Oct	-	-	-	-	237	-
2004-Nov	-	-	-	-	245	-
2004-Dec	-	-	-	-	265	-
2005-Jan	-	-	-	-	270	-
2005-Feb	-	-	-	-	281	-
2005-Mar	-	-	-	-	290	-
2005-Apr	-	-	-	-	300	-
2005-May	-	-	-	-	308	-
2005-Jun	241	-	-	-	336	241
2005-Jul	-	-	-	-	-	245
2005-Aug	-	-	-	-	-	268
2005-Sep	-	-	-	-	-	284
2005-Oct	-	-	-	-	-	294
2005-Nov	-	-	-	-	-	315
2005-Dec	-	-	-	-	-	332
2006-Jan	-	-	-	-	-	343
2006-Feb	-	-	-	-	-	358
2006-Mar	-	-	-	-	-	379
2006-Apr	-	-	-	-	-	394
2006-May	-	-	-	-	-	416
2006-Jun	281	-	-	-	-	433

Many of the third parties to U.S.- U.K. transactions reside in the euro area, and we see in Figure 3a that the actual euro area positions, as measured by the surveys, are almost always higher than the estimates based on transactions data.

Figure 3a. Euro-area holdings of U.S. long-term securities
Estimated monthly positions using adjusted net flows with valuation changes in billions of dollars

Data for Figure 3a immediately follows.

Values constructed from periodic surveys of holdings, adjusted monthly net transactions and valuation adjustments. Includes Channel Islands and the Isle of Man.

Data for Figure 3a - Treasury bonds

Date	Survey value	Survey date 2000-Mar	Survey date 2002-Jun	Survey date 2003-Jun	Survey date 2004-Jun	Survey date 2005-Jun
2000-Mar	149	149	-	-	-	-
2000-Apr	-	149	-	-	-	-
2000-May	-	146	-	-	-	-
2000-Jun	-	148	-	-	-	-
2000-Jul	-	148	-	-	-	-
2000-Aug	-	150	-	-	-	-
2000-Sep	-	150	-	-	-	-
2000-Oct	-	152	-	-	-	-
2000-Nov	-	151	-	-	-	-
2000-Dec	-	149	-	-	-	-
2001-Jan	-	144	-	-	-	-
2001-Feb	-	146	-	-	-	-
2001-Mar	-	142	-	-	-	-
2001-Apr	-	139	-	-	-	-
2001-May	-	139	-	-	-	-
2001-Jun	-	140	-	-	-	-
2001-Jul	-	142	-	-	-	-
2001-Aug	-	147	-	-	-	-
2001-Sep	-	142	-	-	-	-
2001-Oct	-	142	-	-	-	-
2001-Nov	-	136	-	-	-	-
2001-Dec	-	132	-	-	-	-
2002-Jan	-	125	-	-	-	-
2002-Feb	-	127	-	-	-	-
2002-Mar	-	125	-	-	-	-
2002-Apr	-	125	-	-	-	-
2002-May	-	118	-	-	-	-
2002-Jun	135	115	135	-	-	-
2002-Jul	-	-	133	-	-	-
2002-Aug	-	-	133	-	-	-
2002-Sep	-	-	137	-	-	-
2002-Oct	-	-	129	-	-	-
2002-Nov	-	-	133	-	-	-
2002-Dec	-	-	141	-	-	-
2003-Jan	-	-	138	-	-	-
2003-Feb	-	-	137	-	-	-
2003-Mar	-	-	132	-	-	-
2003-Apr	-	-	140	-	-	-
2003-May	-	-	146	-	-	-
2003-Jun	150	-	151	150	-	-
2003-Jul	-	-	-	148	-	-
2003-Aug	-	-	-	153	-	-
2003-Sep	-	-	-	155	-	-
2003-Oct	-	-	-	148	-	-
2003-Nov	-	-	-	149	-	-
2003-Dec	-	-	-	140	-	-
2004-Jan	-	-	-	143	-	-
2004-Feb	-	-	-	137	-	-
2004-Mar	-	-	-	137	-	-
2004-Apr	-	-	-	140	-	-
2004-May	-	-	-	138	-	-
2004-Jun	154	-	-	136	154	-
2004-Jul	-	-	-	-	159	-
2004-Aug	-	-	-	-	150	-
2004-Sep	-	-	-	-	151	-
2004-Oct	-	-	-	-	152	-
2004-Nov	-	-	-	-	152	-
2004-Dec	-	-	-	-	153	-
2005-Jan	-	-	-	-	159	-
2005-Feb	-	-	-	-	161	-
2005-Mar	-	-	-	-	161	-
2005-Apr	-	-	-	-	166	-
2005-May	-	-	-	-	175	-
2005-Jun	156	-	-	-	168	156
2005-Jul	-	-	-	-	-	158
2005-Aug	-	-	-	-	-	161
2005-Sep	-	-	-	-	-	161
2005-Oct	-	-	-	-	-	164
2005-Nov	-	-	-	-	-	164
2005-Dec	-	-	-	-	-	168
2006-Jan	-	-	-	-	-	166
2006-Feb	-	-	-	-	-	163
2006-Mar	-	-	-	-	-	158
2006-Apr	-	-	-	-	-	156
2006-May	-	-	-	-	-	157
2006-Jun	170	-	-	-	-	159

Data for Figure 3a - Corporate bonds

Date	Survey value	Survey date 2000-Mar	Survey date 2002-Jun	Survey date 2003-Jun	Survey date 2004-Jun	Survey date 2005-Jun
2000-Mar	115	115	-	-	-	-
2000-Apr	-	115	-	-	-	-
2000-May	-	116	-	-	-	-
2000-Jun	-	120	-	-	-	-
2000-Jul	-	121	-	-	-	-
2000-Aug	-	123	-	-	-	-
2000-Sep	-	123	-	-	-	-
2000-Oct	-	123	-	-	-	-
2000-Nov	-	126	-	-	-	-
2000-Dec	-	131	-	-	-	-
2001-Jan	-	136	-	-	-	-
2001-Feb	-	138	-	-	-	-
2001-Mar	-	140	-	-	-	-
2001-Apr	-	142	-	-	-	-
2001-May	-	145	-	-	-	-
2001-Jun	-	147	-	-	-	-
2001-Jul	-	151	-	-	-	-
2001-Aug	-	153	-	-	-	-
2001-Sep	-	151	-	-	-	-
2001-Oct	-	156	-	-	-	-
2001-Nov	-	155	-	-	-	-
2001-Dec	-	154	-	-	-	-
2002-Jan	-	156	-	-	-	-
2002-Feb	-	157	-	-	-	-
2002-Mar	-	157	-	-	-	-
2002-Apr	-	158	-	-	-	-
2002-May	-	161	-	-	-	-
2002-Jun	245	164	245	-	-	-
2002-Jul	-	-	245	-	-	-
2002-Aug	-	-	249	-	-	-
2002-Sep	-	-	252	-	-	-
2002-Oct	-	-	248	-	-	-
2002-Nov	-	-	250	-	-	-
2002-Dec	-	-	256	-	-	-
2003-Jan	-	-	257	-	-	-
2003-Feb	-	-	262	-	-	-
2003-Mar	-	-	262	-	-	-
2003-Apr	-	-	265	-	-	-
2003-May	-	-	273	-	-	-
2003-Jun	363	-	273	363	-	-
2003-Jul	-	-	-	350	-	-
2003-Aug	-	-	-	351	-	-
2003-Sep	-	-	-	361	-	-
2003-Oct	-	-	-	357	-	-
2003-Nov	-	-	-	357	-	-
2003-Dec	-	-	-	362	-	-
2004-Jan	-	-	-	366	-	-
2004-Feb	-	-	-	371	-	-
2004-Mar	-	-	-	377	-	-
2004-Apr	-	-	-	365	-	-
2004-May	-	-	-	363	-	-
2004-Jun	545	-	-	366	545	-
2004-Jul	-	-	-	-	552	-
2004-Aug	-	-	-	-	565	-
2004-Sep	-	-	-	-	572	-
2004-Oct	-	-	-	-	576	-
2004-Nov	-	-	-	-	571	-
2004-Dec	-	-	-	-	578	-
2005-Jan	-	-	-	-	582	-
2005-Feb	-	-	-	-	579	-
2005-Mar	-	-	-	-	572	-
2005-Apr	-	-	-	-	578	-
2005-May	-	-	-	-	584	-
2005-Jun	651	-	-	-	588	651
2005-Jul	-	-	-	-	-	644
2005-Aug	-	-	-	-	-	650
2005-Sep	-	-	-	-	-	643
2005-Oct	-	-	-	-	-	638
2005-Nov	-	-	-	-	-	641
2005-Dec	-	-	-	-	-	647
2006-Jan	-	-	-	-	-	644
2006-Feb	-	-	-	-	-	644
2006-Mar	-	-	-	-	-	635
2006-Apr	-	-	-	-	-	632
2006-May	-	-	-	-	-	630
2006-Jun	745	-	-	-	-	629

Unfortunately, we do not have a data source that would permit us to adjust the transactions data for this financial-center effect.

Thus, adjusting for omitted transactions and valuation changes does not completely eliminate the discrepancy between positions as reported in the periodic surveys and as constructed using monthly transactions data, at either the country or the aggregate level. We assume that the surveys accurately measure securities positions as of the survey date; this implies that the sum of the observed, adjusted net transactions, corrected for valuation changes, is in error by the amount of the gap between the survey and the sum of net transactions.⁵ This gap is assumed to represent the financial center effect discussed above, as well as unknown errors and omissions in the monthly transactions data of current S-form reporters, and transactions conducted by entities that have not yet been identified as prospective reporters. In addition, the gap may be due to various measurement and approximation errors in the construction of the prices used to calculate the valuation adjustments, and to transactions costs, which are included in reported transactions but not in survey positions.

3 Estimating monthly positions between surveys

As noted in the introduction, there are a number of reasons why it would be useful to construct a series of estimated monthly positions for the periods between the benchmark surveys, and more particularly, for the period from the latest survey to the most current monthly transactions data. As we have seen, even after making a number of needed adjustments, the monthly data remain noticeably inconsistent with the survey data. For estimates between surveys, simply extending a given survey value using adjusted transactions and valuation changes is unsatisfactory because the resulting positions are inconsistent with the known values of the next survey. And given this observed discrepancy between survey values, it should be possible to do better in estimating current holdings than simply extending forward from the last survey value using net transactions and valuation changes.

Our basic problem, then, is to distribute the known error observed when a new survey is conducted across the months between survey dates so as to generate a more accurate set of monthly position estimates. When applied country by country, this approach will generate estimates that correct for financial center bias and more accurately reflect holdings of residents of the country in question. Of course, we do not know when during the inter-survey period the measurement errors occurred; we only know the size of the cumulative error as observed on the new survey date.

Beginning with an initial survey position, an estimate of the position at a future date can be constructed as

(1) $\hat{S}_{t} =S_{0} (1+\hat{\pi}_{0,t} )+\sum\limits_{i=1}^{t} {\hat {N}_{i} } (1+\hat{\pi}_{i,t} )$

where $S_{0}$ is the latest survey value for a given country, security, and holder; $\hat{S}_{t}$ is the estimated position at time >0, $\{\hat {N}_{i} \}$ is the sequence of net flows from time 1 to t and $\hat{\pi }_{i,t}$ is the rate of increase in the price of security S over the period i to t, with $\pi_{i,i} =0$ .⁶

We assume that $S_{0}$ is actually known; $\hat{N}_{i}$ and $\hat{\pi}_{i,t}$ are observed with error and $\hat{S}_{t}$ is an estimate. When , the date of the next survey, $S_{T}$ is known, and we can define the "gap" between the actual and the estimated survey positions:

(2) $G_{T} =S_{T} -\hat{S}_{T}$

We model the measurement errors in security prices $(\varepsilon_{t} )$ as

(3) $(1+\pi_{t} )=(1+\hat{\pi}_{t} )(1+\varepsilon_{t} )$

where $\varepsilon_{t}$ is the multiplicative error in observing the true monthly valuation change, $\pi_{t}$ .

We model transactions costs and other errors in net transactions as

(4) $N_{t} =(1+\beta_{t} )\hat{N}_{t}$

where $\beta_{t}$ includes both effects in multiplicative form. (We assume that transactions costs and other measurement errors are equal percentages of both purchases and sales, so that $\beta_{t}$ may be applied directly to net transactions.) For the results to be plausible we require $\hat{S}_{t} \ge 0$ for all , something that is by no means true in the data.⁷ We impose nonnegativity by adjusting $\hat{N}_{t}$ by a quantity just large enough to bring $\hat{S}_{t}$ to zero in any month in which it would otherwise be negative.

A little algebra (see Appendix C) shows that we can write the estimated position $\tilde{S}_{t}$ as the constructed position ( $\hat{S}_{t}$ ) plus the share of the final gap allocated to month , discounted for actual valuation changes from time to :

(5) $\tilde{S}_{t} =\tilde{S}_{t-1} (1+\hat{\pi}_{t} )+\hat{N}_{t} +\frac{\lambda_{t} G_{T} }{\pi_{t,T} }$

where

(6) $\lambda_{t} =\lambda\left( {\hat{S}_{i} ,\hat{\pi}_{i} ,\hat{N}_{i} ,\varepsilon_{i} ,\beta_{i} } \right) ,\mbox{ }i=0,T,\mbox{ }\sum{\lambda_{t} } =1.0$

and it can be shown that $\tilde{S}_{t}$ and that the law of motion (equation (1)) holds for $\tilde{S}_{t}$ , so that the sequence ,t = 1,T, is "survey consistent" in the sense of TWW.

The difficulty, of course, is that equation (6) includes the unobserved measurement errors $\varepsilon_{t}$ and $\beta_{t}$ , about which we have essentially no prior information.⁸ We simply don't know whether these errors are serially correlated or not, whether they have zero mean or not, or what their comparative magnitudes might be. In this situation any solution is likely to be almost entirely ad hoc. The approach we follow here is to assume that the measurement errors are uniformly distributed throughout the period between surveys, i.e. that $\varepsilon_{t} =\varepsilon$ and $\beta_{t} =\beta$ . We then assume what we hope are plausible values for $\varepsilon$ and $\beta$ , and perform some limited sensitivity testing.⁹

Equation (5) describes a path for the position in security that obeys the law of motion for the position and is equal to the reported survey positions at time 0 and time .

In brief, we extrapolate the time 0 survey position forward using observed flow data and compute the residual vis-à-vis the reported survey at time . This residual is then distributed across time periods according to each period's share in total net transactions, and each residual share is then discounted by the appropriate inflation rate. The estimates will then match reported surveys by construction, and will be consistent with both endpoints, with the reported monthly transactions, and with measured changes in asset prices.¹⁰

Figure 4 presents our monthly estimated positions for agency bonds held by all foreigners (cf. Figure 1).¹¹ As discussed above, the path of the estimated positions coincides with the survey values for years in which surveys were taken. In recent years, as the time between surveys has been reduced, the difference between positions estimated with unadjusted net flows and those estimated with the correction has also tended to fall. Figure 4a shows the same series from 2003, in more detail.

Figure 4. Total foreign holdings of U.S. agency bonds
Estimated monthly positions using adjusted net flows in billions of dollars

Data for Figure 4 immediately follows.

Values constructed from periodic surveys of holdings and adjusted monthly net transactions.

Data for Figure 4

Date	Survey value	Survey date 1994-Dec: survey consistent estimated positions	Survey date 1994-Dec: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2002-Jun: adj. net flows	Survey date 2003-Jun: adj. net flows	Survey date 2004-Jun: adj. net flows	Survey date 2005-Jun: adj. net flows
1994-Dec	107	107	107	-	-	-	-	-
1995-Jan	-	109	111	-	-	-	-	-
1995-Feb	-	111	114	-	-	-	-	-
1995-Mar	-	112	117	-	-	-	-	-
1995-Apr	-	113	120	-	-	-	-	-
1995-May	-	117	126	-	-	-	-	-
1995-Jun	-	118	128	-	-	-	-	-
1995-Jul	-	118	130	-	-	-	-	-
1995-Aug	-	121	135	-	-	-	-	-
1995-Sep	-	125	141	-	-	-	-	-
1995-Oct	-	125	143	-	-	-	-	-
1995-Nov	-	128	149	-	-	-	-	-
1995-Dec	-	126	148	-	-	-	-	-
1996-Jan	-	126	151	-	-	-	-	-
1996-Feb	-	125	151	-	-	-	-	-
1996-Mar	-	122	149	-	-	-	-	-
1996-Apr	-	120	149	-	-	-	-	-
1996-May	-	120	151	-	-	-	-	-
1996-Jun	-	123	157	-	-	-	-	-
1996-Jul	-	126	161	-	-	-	-	-
1996-Aug	-	127	164	-	-	-	-	-
1996-Sep	-	130	170	-	-	-	-	-
1996-Oct	-	134	176	-	-	-	-	-
1996-Nov	-	141	185	-	-	-	-	-
1996-Dec	-	139	186	-	-	-	-	-
1997-Jan	-	141	189	-	-	-	-	-
1997-Feb	-	143	194	-	-	-	-	-
1997-Mar	-	143	195	-	-	-	-	-
1997-Apr	-	148	203	-	-	-	-	-
1997-May	-	149	206	-	-	-	-	-
1997-Jun	-	150	210	-	-	-	-	-
1997-Jul	-	156	220	-	-	-	-	-
1997-Aug	-	160	224	-	-	-	-	-
1997-Sep	-	160	228	-	-	-	-	-
1997-Oct	-	167	239	-	-	-	-	-
1997-Nov	-	164	238	-	-	-	-	-
1997-Dec	-	165	242	-	-	-	-	-
1998-Jan	-	168	248	-	-	-	-	-
1998-Feb	-	172	254	-	-	-	-	-
1998-Mar	-	179	264	-	-	-	-	-
1998-Apr	-	184	271	-	-	-	-	-
1998-May	-	185	276	-	-	-	-	-
1998-Jun	-	190	284	-	-	-	-	-
1998-Jul	-	189	285	-	-	-	-	-
1998-Aug	-	193	293	-	-	-	-	-
1998-Sep	-	193	298	-	-	-	-	-
1998-Oct	-	183	290	-	-	-	-	-
1998-Nov	-	190	299	-	-	-	-	-
1998-Dec	-	194	306	-	-	-	-	-
1999-Jan	-	200	315	-	-	-	-	-
1999-Feb	-	197	312	-	-	-	-	-
1999-Mar	-	205	323	-	-	-	-	-
1999-Apr	-	213	334	-	-	-	-	-
1999-May	-	214	336	-	-	-	-	-
1999-Jun	-	215	339	-	-	-	-	-
1999-Jul	-	217	344	-	-	-	-	-
1999-Aug	-	219	347	-	-	-	-	-
1999-Sep	-	227	359	-	-	-	-	-
1999-Oct	-	233	368	-	-	-	-	-
1999-Nov	-	236	374	-	-	-	-	-
1999-Dec	-	236	375	-	-	-	-	-
2000-Jan	-	237	378	-	-	-	-	-
2000-Feb	-	249	394	-	-	-	-	-
2000-Mar	261	261	410	261	-	-	-	-
2000-Apr	-	264	-	269	-	-	-	-
2000-May	-	272	-	280	-	-	-	-
2000-Jun	-	278	-	291	-	-	-	-
2000-Jul	-	284	-	302	-	-	-	-
2000-Aug	-	299	-	321	-	-	-	-
2000-Sep	-	311	-	339	-	-	-	-
2000-Oct	-	325	-	358	-	-	-	-
2000-Nov	-	342	-	381	-	-	-	-
2000-Dec	-	352	-	398	-	-	-	-
2001-Jan	-	363	-	415	-	-	-	-
2001-Feb	-	368	-	426	-	-	-	-
2001-Mar	-	381	-	446	-	-	-	-
2001-Apr	-	386	-	457	-	-	-	-
2001-May	-	387	-	463	-	-	-	-
2001-Jun	-	396	-	479	-	-	-	-
2001-Jul	-	409	-	500	-	-	-	-
2001-Aug	-	416	-	515	-	-	-	-
2001-Sep	-	424	-	531	-	-	-	-
2001-Oct	-	451	-	567	-	-	-	-
2001-Nov	-	448	-	569	-	-	-	-
2001-Dec	-	444	-	571	-	-	-	-
2002-Jan	-	445	-	580	-	-	-	-
2002-Feb	-	446	-	589	-	-	-	-
2002-Mar	-	449	-	596	-	-	-	-
2002-Apr	-	473	-	631	-	-	-	-
2002-May	-	485	-	651	-	-	-	-
2002-Jun	492	492	-	668	492	-	-	-
2002-Jul	-	503	-	-	508	-	-	-
2002-Aug	-	517	-	-	527	-	-	-
2002-Sep	-	533	-	-	550	-	-	-
2002-Oct	-	542	-	-	564	-	-	-
2002-Nov	-	544	-	-	571	-	-	-
2002-Dec	-	556	-	-	589	-	-	-
2003-Jan	-	566	-	-	605	-	-	-
2003-Feb	-	569	-	-	614	-	-	-
2003-Mar	-	570	-	-	621	-	-	-
2003-Apr	-	575	-	-	633	-	-	-
2003-May	-	595	-	-	659	-	-	-
2003-Jun	586	586	-	-	656	586	-	-
2003-Jul	-	567	-	-	-	570	-	-
2003-Aug	-	563	-	-	-	570	-	-
2003-Sep	-	564	-	-	-	574	-	-
2003-Oct	-	559	-	-	-	573	-	-
2003-Nov	-	560	-	-	-	577	-	-
2003-Dec	-	571	-	-	-	591	-	-
2004-Jan	-	593	-	-	-	616	-	-
2004-Feb	-	611	-	-	-	638	-	-
2004-Mar	-	605	-	-	-	635	-	-
2004-Apr	-	607	-	-	-	640	-	-
2004-May	-	612	-	-	-	649	-	-
2004-Jun	619	619	-	-	-	659	619	-
2004-Jul	-	638	-	-	-	-	638	-
2004-Aug	-	660	-	-	-	-	660	-
2004-Sep	-	663	-	-	-	-	663	-
2004-Oct	-	684	-	-	-	-	684	-
2004-Nov	-	700	-	-	-	-	701	-
2004-Dec	-	724	-	-	-	-	725	-
2005-Jan	-	746	-	-	-	-	746	-
2005-Feb	-	750	-	-	-	-	751	-
2005-Mar	-	748	-	-	-	-	749	-
2005-Apr	-	756	-	-	-	-	757	-
2005-May	-	777	-	-	-	-	778	-
2005-Jun	791	791	-	-	-	-	792	791
2005-Jul	-	813	-	-	-	-	-	811
2005-Aug	-	829	-	-	-	-	-	826
2005-Sep	-	834	-	-	-	-	-	829
2005-Oct	-	855	-	-	-	-	-	849
2005-Nov	-	862	-	-	-	-	-	854
2005-Dec	-	873	-	-	-	-	-	863
2006-Jan	-	898	-	-	-	-	-	886
2006-Feb	-	924	-	-	-	-	-	911
2006-Mar	-	928	-	-	-	-	-	913
2006-Apr	-	939	-	-	-	-	-	922
2006-May	-	968	-	-	-	-	-	949
2006-Jun	984	984	-	-	-	-	-	964

Figure 4a. Total foreign holdings of U.S. agency bonds
Estimated monthly positions using adjusted net flows in billions of dollars

Data for Figure 4a immediately follows.

Values constructed from periodic surveys of holdings and adjusted monthly net transactions.

Data for Figure 4a

Date	Survey value	Survey date 2003-Jan: survey-consistent estimated positions	Survey date 2004-Jun: adj. net flows	Survey date 2003-Jun: adj. net flows	Survey date 2005-Jun: adj. net flows
2003-Jan	-	566	-	-	-
2003-Feb	-	569	-	-	-
2003-Mar	-	570	-	-	-
2003-Apr	-	575	-	-	-
2003-May	-	595	-	-	-
2003-Jun	586	586	586	-	-
2003-Jul	-	567	570	-	-
2003-Aug	-	563	570	-	-
2003-Sep	-	564	574	-	-
2003-Oct	-	559	573	-	-
2003-Nov	-	560	577	-	-
2003-Dec	-	571	591	-	-
2004-Jan	-	593	616	-	-
2004-Feb	-	611	638	-	-
2004-Mar	-	605	635	-	-
2004-Apr	-	607	640	-	-
2004-May	-	612	649	-	-
2004-Jun	619	619	659	619	-
2004-Jul	-	638	-	638	-
2004-Aug	-	660	-	660	-
2004-Sep	-	663	-	663	-
2004-Oct	-	684	-	684	-
2004-Nov	-	700	-	701	-
2004-Dec	-	724	-	725	-
2005-Jan	-	746	-	746	-
2005-Feb	-	750	-	751	-
2005-Mar	-	748	-	749	-
2005-Apr	-	756	-	757	-
2005-May	-	777	-	778	-
2005-Jun	791	791	-	792	791
2005-Jul	-	813	-	-	811
2005-Aug	-	829	-	-	826
2005-Sep	-	834	-	-	829
2005-Oct	-	855	-	-	849
2005-Nov	-	862	-	-	854
2005-Dec	-	873	-	-	863
2006-Jan	-	898	-	-	886
2006-Feb	-	924	-	-	911
2006-Mar	-	928	-	-	913
2006-Apr	-	939	-	-	922
2006-May	-	968	-	-	949
2006-Jun	984	984	-	-	964

In contrast, the difference for treasury bonds held by agents in the U.K. is dramatic, as shown in Figure 5. By distributing the gaps across the period between surveys, the large financial center bias is eliminated:

Figure 5. U.K. holdings of U.S. treasury bonds
Estimated monthly positions using adjusted net flows in billions of dollars

Data for Figure 5 immediately follows.

Values constructed from periodic surveys of holdings and adjusted monthly net transactions.

Data for Figure 5

Date	Survey value	Survey date 1994-Dec: survey consistent estimated positions	Survey date 1994-Dec: adj. net flows	Survey date 2000-Mar: adj. net flows	Survey date 2002-Jun: adj. net flows	Survey date 2003-Jun: adj. net flows	Survey date 2004-Jun: adj. net flows	Survey date 2005-Jun: adj. net flows
1994-Dec	11	11	11	-	-	-	-	-
1995-Jan	-	13	13	-	-	-	-	-
1995-Feb	-	15	16	-	-	-	-	-
1995-Mar	-	15	16	-	-	-	-	-
1995-Apr	-	16	18	-	-	-	-	-
1995-May	-	16	18	-	-	-	-	-
1995-Jun	-	17	20	-	-	-	-	-
1995-Jul	-	17	21	-	-	-	-	-
1995-Aug	-	19	23	-	-	-	-	-
1995-Sep	-	21	26	-	-	-	-	-
1995-Oct	-	22	27	-	-	-	-	-
1995-Nov	-	24	30	-	-	-	-	-
1995-Dec	-	23	30	-	-	-	-	-
1996-Jan	-	23	31	-	-	-	-	-
1996-Feb	-	25	33	-	-	-	-	-
1996-Mar	-	24	33	-	-	-	-	-
1996-Apr	-	23	33	-	-	-	-	-
1996-May	-	24	35	-	-	-	-	-
1996-Jun	-	25	37	-	-	-	-	-
1996-Jul	-	26	39	-	-	-	-	-
1996-Aug	-	26	40	-	-	-	-	-
1996-Sep	-	27	41	-	-	-	-	-
1996-Oct	-	27	43	-	-	-	-	-
1996-Nov	-	27	44	-	-	-	-	-
1996-Dec	-	25	43	-	-	-	-	-
1997-Jan	-	25	44	-	-	-	-	-
1997-Feb	-	26	46	-	-	-	-	-
1997-Mar	-	27	47	-	-	-	-	-
1997-Apr	-	28	49	-	-	-	-	-
1997-May	-	28	51	-	-	-	-	-
1997-Jun	-	27	51	-	-	-	-	-
1997-Jul	-	26	52	-	-	-	-	-
1997-Aug	-	27	53	-	-	-	-	-
1997-Sep	-	26	54	-	-	-	-	-
1997-Oct	-	26	55	-	-	-	-	-
1997-Nov	-	24	54	-	-	-	-	-
1997-Dec	-	22	54	-	-	-	-	-
1998-Jan	-	21	54	-	-	-	-	-
1998-Feb	-	21	55	-	-	-	-	-
1998-Mar	-	25	60	-	-	-	-	-
1998-Apr	-	26	62	-	-	-	-	-
1998-May	-	25	63	-	-	-	-	-
1998-Jun	-	27	66	-	-	-	-	-
1998-Jul	-	28	69	-	-	-	-	-
1998-Aug	-	29	71	-	-	-	-	-
1998-Sep	-	31	76	-	-	-	-	-
1998-Oct	-	35	80	-	-	-	-	-
1998-Nov	-	39	86	-	-	-	-	-
1998-Dec	-	39	88	-	-	-	-	-
1999-Jan	-	38	88	-	-	-	-	-
1999-Feb	-	37	87	-	-	-	-	-
1999-Mar	-	37	89	-	-	-	-	-
1999-Apr	-	36	90	-	-	-	-	-
1999-May	-	36	90	-	-	-	-	-
1999-Jun	-	37	92	-	-	-	-	-
1999-Jul	-	35	92	-	-	-	-	-
1999-Aug	-	35	93	-	-	-	-	-
1999-Sep	-	35	95	-	-	-	-	-
1999-Oct	-	35	96	-	-	-	-	-
1999-Nov	-	34	96	-	-	-	-	-
1999-Dec	-	32	95	-	-	-	-	-
2000-Jan	-	30	95	-	-	-	-	-
2000-Feb	-	33	99	-	-	-	-	-
2000-Mar	36	36	104	36	-	-	-	-
2000-Apr	-	35	-	36	-	-	-	-
2000-May	-	36	-	39	-	-	-	-
2000-Jun	-	34	-	39	-	-	-	-
2000-Jul	-	31	-	38	-	-	-	-
2000-Aug	-	31	-	40	-	-	-	-
2000-Sep	-	37	-	47	-	-	-	-
2000-Oct	-	39	-	52	-	-	-	-
2000-Nov	-	41	-	56	-	-	-	-
2000-Dec	-	42	-	60	-	-	-	-
2001-Jan	-	43	-	64	-	-	-	-
2001-Feb	-	43	-	67	-	-	-	-
2001-Mar	-	45	-	72	-	-	-	-
2001-Apr	-	44	-	73	-	-	-	-
2001-May	-	44	-	76	-	-	-	-
2001-Jun	-	48	-	83	-	-	-	-
2001-Jul	-	49	-	88	-	-	-	-
2001-Aug	-	48	-	91	-	-	-	-
2001-Sep	-	49	-	96	-	-	-	-
2001-Oct	-	55	-	107	-	-	-	-
2001-Nov	-	54	-	108	-	-	-	-
2001-Dec	-	48	-	106	-	-	-	-
2002-Jan	-	46	-	108	-	-	-	-
2002-Feb	-	42	-	109	-	-	-	-
2002-Mar	-	40	-	109	-	-	-	-
2002-Apr	-	45	-	119	-	-	-	-
2002-May	-	41	-	120	-	-	-	-
2002-Jun	40	40	-	123	40	-	-	-
2002-Jul	-	37	-	-	39	-	-	-
2002-Aug	-	36	-	-	41	-	-	-
2002-Sep	-	38	-	-	46	-	-	-
2002-Oct	-	40	-	-	52	-	-	-
2002-Nov	-	38	-	-	54	-	-	-
2002-Dec	-	36	-	-	55	-	-	-
2003-Jan	-	41	-	-	64	-	-	-
2003-Feb	-	39	-	-	67	-	-	-
2003-Mar	-	37	-	-	69	-	-	-
2003-Apr	-	37	-	-	74	-	-	-
2003-May	-	36	-	-	78	-	-	-
2003-Jun	29	29	-	-	75	29	-	-
2003-Jul	-	27	-	-	-	29	-	-
2003-Aug	-	25	-	-	-	29	-	-
2003-Sep	-	23	-	-	-	29	-	-
2003-Oct	-	20	-	-	-	27	-	-
2003-Nov	-	19	-	-	-	28	-	-
2003-Dec	-	18	-	-	-	30	-	-
2004-Jan	-	23	-	-	-	36	-	-
2004-Feb	-	25	-	-	-	41	-	-
2004-Mar	-	18	-	-	-	36	-	-
2004-Apr	-	21	-	-	-	41	-	-
2004-May	-	23	-	-	-	45	-	-
2004-Jun	24	24	-	-	-	49	24	-
2004-Jul	-	26	-	-	-	-	28	-
2004-Aug	-	29	-	-	-	-	33	-
2004-Sep	-	31	-	-	-	-	37	-
2004-Oct	-	32	-	-	-	-	41	-
2004-Nov	-	35	-	-	-	-	47	-
2004-Dec	-	38	-	-	-	-	54	-
2005-Jan	-	39	-	-	-	-	58	-
2005-Feb	-	41	-	-	-	-	65	-
2005-Mar	-	41	-	-	-	-	68	-
2005-Apr	-	34	-	-	-	-	66	-
2005-May	-	30	-	-	-	-	66	-
2005-Jun	28	28	-	-	-	-	68	28
2005-Jul	-	32	-	-	-	-	-	34
2005-Aug	-	32	-	-	-	-	-	38
2005-Sep	-	33	-	-	-	-	-	43
2005-Oct	-	34	-	-	-	-	-	47
2005-Nov	-	31	-	-	-	-	-	47
2005-Dec	-	30	-	-	-	-	-	51
2006-Jan	-	29	-	-	-	-	-	53
2006-Feb	-	30	-	-	-	-	-	59
2006-Mar	-	31	-	-	-	-	-	65
2006-Apr	-	30	-	-	-	-	-	68
2006-May	-	30	-	-	-	-	-	73
2006-Jun	31	31	-	-	-	-	-	80

In constructing these estimates, we assume that both $\varepsilon$ and $\beta$ are equal to .01, implying that the measurement errors are on the order of 1% of monthly transactions and that errors in measuring valuation effects and net flows are of the same magnitude. These assumed values do not have any real empirical basis; rather they are (we believe) plausible values and relatively neutral assumptions that provide plausible results. Figures 6 and 6a illustrate the sensitivity of our estimates to these assumptions. In each figure, we show estimated positions for values of $\varepsilon$ and $\beta$ from -.01 to .02. As shown in the two figures, varying the errors over this range has a modest effect in the early part of our sample, when the period between surveys was two years or more. Since the advent of annual surveys in 2003 the estimated positions are essentially insensitive to changes in the error assumptions over this range.

Figure 6. Total foreign holdings of U.S. agency bonds
Estimated monthly positions with different error parameters in billions of dollars

Data for Figure 6 immediately follows.

Values constructed from periodic surveys of holdings and adjusted monthly net transactions.

Data for Figure 6

Date	$\varepsilon$ = -0.01, $\beta$ = -0.01	$\varepsilon$ = 0.01, $\beta$ = 0.01	$\varepsilon$ = 0.02, $\beta$ = 0.02
1994-Jun	106	104	103
1994-Jul	108	107	106
1994-Aug	109	108	107
1994-Sep	106	105	104
1994-Oct	105	104	104
1994-Nov	106	106	106
1994-Dec	107	107	107
1995-Jan	110	109	108
1995-Feb	113	111	110
1995-Mar	115	112	110
1995-Apr	117	113	111
1995-May	121	117	114
1995-Jun	123	118	114
1995-Jul	124	118	113
1995-Aug	128	121	116
1995-Sep	133	125	119
1995-Oct	134	125	119
1995-Nov	138	128	122
1995-Dec	136	126	119
1996-Jan	138	126	119
1996-Feb	137	125	117
1996-Mar	134	122	114
1996-Apr	133	120	111
1996-May	134	120	112
1996-Jun	138	123	114
1996-Jul	140	125	116
1996-Aug	142	127	117
1996-Sep	147	130	121
1996-Oct	151	134	124
1996-Nov	158	140	130
1996-Dec	157	139	129
1997-Jan	159	141	130
1997-Feb	163	143	132
1997-Mar	162	143	132
1997-Apr	168	148	137
1997-May	170	149	138
1997-Jun	171	150	138
1997-Jul	178	156	144
1997-Aug	181	159	148
1997-Sep	182	160	148
1997-Oct	190	167	155
1997-Nov	187	164	152
1997-Dec	188	165	153
1998-Jan	191	168	156
1998-Feb	195	172	161
1998-Mar	202	179	168
1998-Apr	207	184	173
1998-May	208	185	174
1998-Jun	213	190	179
1998-Jul	211	189	178
1998-Aug	215	193	182
1998-Sep	215	193	182
1998-Oct	204	183	173
1998-Nov	210	189	180
1998-Dec	214	194	184
1999-Jan	219	200	191
1999-Feb	215	196	188
1999-Mar	223	205	197
1999-Apr	230	213	205
1999-May	230	214	207
1999-Jun	229	215	208
1999-Jul	231	217	212
1999-Aug	231	219	214
1999-Sep	237	227	222
1999-Oct	242	233	229
1999-Nov	244	236	233
1999-Dec	242	236	234
2000-Jan	241	237	235
2000-Feb	251	249	248
2000-Mar	261	261	261
2000-Apr	265	264	264
2000-May	274	272	271
2000-Jun	281	278	276
2000-Jul	289	284	282
2000-Aug	304	299	296
2000-Sep	317	311	308
2000-Oct	332	325	321
2000-Nov	350	342	338
2000-Dec	360	352	347
2001-Jan	372	362	358
2001-Feb	378	368	362
2001-Mar	391	381	375
2001-Apr	397	386	381
2001-May	397	386	381
2001-Jun	407	396	391
2001-Jul	420	409	403
2001-Aug	427	416	411
2001-Sep	434	423	418
2001-Oct	461	451	445
2001-Nov	457	448	443
2001-Dec	453	444	440
2002-Jan	453	445	441
2002-Feb	453	446	443
2002-Mar	454	449	446
2002-Apr	477	473	471
2002-May	487	485	484
2002-Jun	492	492	492
2002-Jul	503	503	502
2002-Aug	518	517	516
2002-Sep	535	533	532
2002-Oct	544	542	541
2002-Nov	546	544	543
2002-Dec	558	556	555

Figure 6a. Total foreign holdings of U.S. agency bonds
Estimated monthly positions with different error parameters in billions of dollars

Data for Figure 6a immediately follows.

Values constructed from periodic surveys of holdings and adjusted monthly net transactions.

Data for Figure 6a

Date	$\varepsilon$ = -0.01, $\beta$ = -0.01	$\varepsilon$ = 0.01, $\beta$ = 0.01	$\varepsilon$ = 0.02, $\beta$ = 0.02
2002-Jan	453	445	441
2002-Feb	453	446	443
2002-Mar	454	449	446
2002-Apr	477	473	471
2002-May	487	485	484
2002-Jun	492	492	492
2002-Jul	503	503	502
2002-Aug	518	517	516
2002-Sep	535	533	532
2002-Oct	544	542	541
2002-Nov	546	544	543
2002-Dec	558	556	555
2003-Jan	568	566	565
2003-Feb	570	568	568
2003-Mar	571	570	569
2003-Apr	577	575	575
2003-May	595	595	594
2003-Jun	586	586	586
2003-Jul	567	567	567
2003-Aug	564	564	563
2003-Sep	565	564	563
2003-Oct	561	559	559
2003-Nov	561	560	560
2003-Dec	572	571	571
2004-Jan	594	593	592
2004-Feb	612	611	610
2004-Mar	606	605	604
2004-Apr	607	607	606
2004-May	613	612	612
2004-Jun	619	619	619
2004-Jul	638	638	638
2004-Aug	660	660	660
2004-Sep	663	663	663
2004-Oct	684	684	684
2004-Nov	700	700	700
2004-Dec	724	724	724
2005-Jan	746	746	746
2005-Feb	750	750	750
2005-Mar	748	748	748
2005-Apr	756	756	756
2005-May	777	777	777
2005-Jun	791	791	791
2005-Jul	812	812	813
2005-Aug	829	829	829
2005-Sep	834	834	834
2005-Oct	854	855	855
2005-Nov	861	862	862
2005-Dec	872	872	873
2006-Jan	896	897	897
2006-Feb	923	923	924
2006-Mar	927	927	927
2006-Apr	938	938	938
2006-May	967	967	968
2006-Jun	984	984	984

4 Estimating positions after the last survey

So far we have assumed that the last date of interest happened to be that of a benchmark survey, so that the problem is confined to constructing position estimates for the period between two surveys. In this case the total cumulative observation error from one survey to the next (the "gap") is known, and our problem is merely to distribute that error to the intermediate observations. In the case where the transactions data extends beyond the last survey (a case of considerable practical importance) we must in effect forecast the measurement error, or gap, out to the end of the transactions data. This turns out to be a difficult empirical problem, and we do not claim to have found a wholly satisfactory solution to it. The crux of the problem is that while the actual survey gaps appear to be very close to pure noise, they nonetheless seem to contain enough information that simply ignoring them in constructing future position estimates would be a mistake. But as we discuss in the next section, our efforts to estimate even the simplest panel regression model of the gaps were largely unsuccessful.¹²

4.1 Summary statistics

We begin our statistical analysis by constructing survey gaps for U.S. liabilities for 84 countries, four security types, and seven survey dates, through the survey of June 2005. (We reserve the newly-released June 2006 survey for out-of-sample testing.) We scale the gaps by the estimated position $\hat{S}_{t}$ to make them comparable across surveys and securities.¹³ Table 1 presents summary statistics for the scaled gaps. The units are simple ratios, so that, for example, the average gap for agency bonds positions, averaged across all countries and all surveys, was 2.8 times the estimated position. It is immediately apparent that there are some extremely large outliers in the gaps that will have to be dealt with somehow. This observation is reinforced by the histograms of the gaps by security type (Figure 7).

Table 1. Summary statistics for scaled gaps, U.S. liabilities to all foreigners by security type and survey date

Survey date	Treasury bonds	Agency bonds	Corporate bonds	Equities	All securities
1989 Survey: mean	1.27	20.76	0.01	0.43	5.52
1989 Survey: minimum	-1.00	-1.00	-1.00	-1.00	-1.00
1989 Survey: maximum	28.73	1547.38	12.34	16.65	1547.38
1989 Survey: std dev	5.57	178.67	1.95	2.31	88.49
1989 Survey: num obs	75	75	76	80	306
1994 Survey: mean	1.90	-0.09	0.13	13.34	4.00
1994 Survey: minimum	-1.00	-1.00	-1.00	-1.00	-1.00
1994 Survey: maximum	114.68	7.55	13.50	936.96	936.96
1994 Survey: std dev	13.59	1.33	1.97	105.65	54.77
1989 Survey: num obs	73	74	73	79	299
2000 Survey: mean	0.14	0.02	-0.10	0.46	0.14
2000 Survey: minimum	-1.00	-1.00	-1.00	-0.97	-1.00
2000 Survey: maximum	6.17	6.88	5.08	33.99	33.99
2000 Survey: std dev	1.11	1.25	1.14	3.93	2.25
2000 Survey: num obs	77	78	72	81	308
2002 Survey: mean	0.18	-0.14	-0.18	0.21	0.02
2002 Survey: minimum	-0.99	-1.00	-1.00	-1.00	-1.00
2002 Survey: maximum	7.30	7.77	3.57	8.51	8.51
2002 Survey: std dev	1.24	1.19	0.70	1.17	1.11
2002 Survey: num obs	79	80	79	81	319
2003 Survey: mean	0.91	-0.14	0.28	0.06	0.28
2003 Survey: minimum	-1.00	-1.00	-1.00	-1.00	-1.00
2003 Survey: maximum	74.95	3.56	42.13	1.72	74.95
2003 Survey: std dev	8.40	0.57	4.76	0.43	4.82
2003 Survey: num obs	80	79	80	82	321
2004 Survey: mean	1.07	0.18	1.90	1.45	1.14
2004 Survey: minimum	-1.00	-1.00	-1.00	-0.55	-1.00
2004 Survey: maximum	73.71	9.56	132.67	82.90	132.67
2004 Survey: std dev	8.24	1.41	14.93	9.26	9.65
2004 Survey: num obs	81	82	79	81	323
2005 Survey: mean	0.08	0.06	-0.03	0.05	0.04
2005 Survey: minimum	-0.88	-1.00	-1.00	-1.00	-1.00
2005 Survey: maximum	1.73	4.36	10.09	2.68	10.09
2005 Survey: std dev	0.53	0.77	1.20	0.47	0.79
2005 Survey: num obs	78	82	81	83	324
All Surveys: mean	0.78	2.82	0.29	2.24	1.55
All Surveys: minimum	-1.00	-1.00	-1.00	-1.00	-1.00
All Surveys: maximum	114.68	1547.38	132.67	936.96	1547.38
All Surveys: std dev	7.06	65.99	6.13	39.67	38.92
All Surveys: num obs	543	550	540	567	2200

Figure 7. Survey Gap Histograms: US Liabilities

Scaled by estimated position. X-axis numbers are the bottom of each cell

Data for Figure 7 immediately follows.

Data for Figure 7

Cell Lower Bound	U.S. Treasuries, All Surveys number of observations)	Agency Bonds, All Surveys (number of observations)	Corporate Bonds, All Surveys (number of observations)	Equities, All Surveys (number of observations)
-0.95	21	28	28	5
-0.90	8	18	19	3
-0.85	7	12	13	6
-0.80	4	12	11	4
-0.75	5	6	8	8
-0.70	7	11	17	2
-0.65	7	15	13	2
-0.60	8	15	17	6
-0.55	8	16	19	12
-0.50	8	10	15	9
-0.45	10	17	17	11
-0.40	9	16	17	9
-0.35	10	13	11	8
-0.30	12	19	13	14
-0.25	14	21	18	20
-0.20	16	22	19	21
-0.15	14	16	16	21
-0.10	23	14	25	18
-0.05	29	10	11	22
0.00	29	19	18	31
0.05	53	70	46	50
0.10	33	14	17	47
0.15	27	19	16	34
0.20	17	12	18	21
0.25	14	8	7	19
0.30	16	6	10	17
0.35	16	11	10	15
0.40	7	11	7	11
0.45	2	9	1	11
0.50	7	5	3	11
0.55	7	7	6	9
0.60	11	4	5	4
0.65	3	5	8	5
0.70	3	2	1	5
0.75	4	4	4	5
0.80	6	2	2	6
0.85	5	1	4	2
0.90	0	2	2	1
0.95	8	3	2	4
1.00	55	45	46	58

Outliers of this magnitude can potentially have a dramatic effect on any empirical model of the gap, and we found even the simplest rules for estimating the gap to be extremely sensitive to the presence or absence of such values. One approach to handling the problem would be to omit observations that are so far from the mean that the hypothesis that they are drawn from the same distribution can be rejected. Unfortunately, as Figure 7 strongly suggests, the sample distributions are clearly non-normal, and as Table 1 shows, the sample standard deviations are so large that even if the distribution were known, this approach would fail to exclude many extremely large values of the gap.¹⁴

For now we fall back on a distinctly ad hoc approach: we exclude from the sample any observations where the measured survey position is more than twice the magnitude of the estimated position. The rationale is that such cases are clearly unusual and appear to represent something other than the "usual" measurement error. While we must nevertheless take these errors into account somehow, we would like to know if the truncated sample reveals any patterns that would be useful in forecasting. Table 2 shows the summary statistics for the scaled gaps with the outliers removed.

Table 2. Summary statistics for scaled gaps: U.S. liabilities to total foreigners, by security type and survey date using truncated data

Survey date	Treasury bonds	Agency bonds	Corporate bonds	Equities	All securities
1989 Survey: mean	0.15	0.13	0.01	0.04	0.09
1989 Survey: minimum	-0.66	-0.66	-0.61	-0.65	-0.66
1989 Survey: maximum	1.99	1.78	1.55	1.43	1.99
1989 Survey: std dev	0.63	0.57	0.53	0.54	0.57
1989 Survey: num obs	60	47	41	67	215
1994 Survey: mean	0.23	0.05	0.08	0.11	0.12
1994 Survey: minimum	-0.65	-0.65	-0.66	-0.66	-0.66
1994 Survey: maximum	1.92	1.64	1.86	1.67	1.92
1994 Survey: std dev	0.56	0.6	0.67	0.49	0.57
1989 Survey: num obs	53	44	49	64	210
2000 Survey: mean	0.12	-0.05	0.04	-0.02	0.02
2000 Survey: minimum	-0.63	-0.65	-0.66	-0.64	-0.66
2000 Survey: maximum	1.96	1.48	1.49	1.96	1.96
2000 Survey: std dev	0.54	0.43	0.62	0.57	0.54
2000 Survey: num obs	60	55	44	66	225
2002 Survey: mean	0.05	-0.11	-0.07	0.11	0.00
2002 Survey: minimum	-0.62	-0.66	-0.64	-0.62	-0.66
2002 Survey: maximum	1.61	1.03	1.70	1.33	1.70
2002 Survey: std dev	0.45	0.43	0.51	0.40	0.45
2002 Survey: num obs	69	57	63	71	260
2003 Survey: mean	-0.01	-0.11	-0.14	0.07	-0.04
2003 Survey: minimum	-0.62	-0.64	-0.66	-0.56	-0.66
2003 Survey: maximum	1.07	1.30	1.62	1.72	1.72
2003 Survey: std dev	0.30	0.33	0.39	0.41	0.37
2003 Survey: num obs	73	70	67	81	291
2004 Survey: mean	0.08	0.04	0.17	0.16	0.11
2004 Survey: minimum	-0.57	-0.62	-0.66	-0.55	-0.66
2004 Survey: maximum	1.63	1.41	1.95	1.41	1.95
2004 Survey: std dev	0.40	0.47	0.51	0.39	0.45
2004 Survey: num obs	67	70	70	75	282
2005 Survey: mean	0.08	0.04	0.17	0.16	0.11
2005 Survey: minimum	-0.57	-0.62	-0.66	-0.55	-0.66
2005 Survey: maximum	1.63	1.41	1.95	1.41	1.95
2005 Survey: std dev	0.40	0.47	0.51	0.39	0.45
2005 Survey: num obs	74	74	72	81	301
All Surveys: mean	0.10	-0.01	0.00	0.07	0.04
All Surveys: minimum	-0.66	-0.66	-0.66	-0.66	-0.66
All Surveys: maximum	1.99	2.00	1.95	1.96	2.00
All Surveys: std dev	0.49	0.48	0.51	0.45	0.48
All Surveys: num obs	456	417	406	505	1784

As would be expected, the standard deviations are much smaller in this sample, although there is still considerable dispersion. A total of 137 observations, out of 2242, were dropped. Comparing by security type, the descriptive statistics indicate that on average, the gaps are somewhat negative for agency bonds (the estimated positions tend to overstate the measured survey positions), slightly positive for Treasuries and stocks, whereas those for corporate bonds tend on average to be zero. For a given security, the gaps also tend to vary in size and sign from year to year.

We also looked for evidence that gaps by country are correlated over time or by security type. Tables 3a and 3b present the correlations over time, for the samples both with and without outliers. For the full sample, there is essentially no correlation at all between one survey year and another. For the truncated sample we observe correlation coefficients on the order of 0.1 - 0.2, suggesting a very modest degree of persistence in the gaps from one survey to the next.

Table 3a. Correlation matrix for gaps by survey years, 1989 - 2005 (full sample, 235 observations)

Survey year	1989	1994	2000	2002	2003	2004	2005
1989	1	-0.006	-0.005	0.053	-0.004	-0.005	-0.026
1994	-	1	-0.001	0.033	-0.006	-0.009	-0.034
2000	-	-	1	-0.002	-0.013	-0.032	0.118
2002	-	-	-	1	-0.018	-0.072	-0.010
2003	-	-	-	-	1	-0.008	-0.016
2004	-	-	-	-	-	1	0.030
2005	-	-	-	-	-	-	1

Table 3b. Correlation matrix for gaps by survey years, 1989 - 2005 (truncated sample, 194 observations)

Survey year	1989	1994	2000	2002	2003	2004	2005
1989	1	-0.063	0.155	0.227	-0.039	-0.062	0.140
1994	-	1	0.106	0.055	-0.059	0.034	-0.023
2000	-	-	1	-0.040	0.116	0.129	-0.087
2002	-	-	-	1	0.315	0.073	0.130
2003	-	-	-	-	1	0.072	-0.023
2004	-	-	-	-	-	1	0.177
2005	-	-	-	-	-	-	1

We find a similar pattern for the correlations across securities - essentially nothing for the full sample, and coefficients on the order of 0.1 - 0.2 for the truncated sample. This result suggests, plausibly enough, that there is some slight tendency for large gaps in a given country for one security to be associated with large gaps in the other securities.

Table 4a. Correlation matrix for gaps by security type (full sample, 476 observations)

Security type	Treasuries	Agencies	Corporates	Equities
Treasury bonds	1	0.026	0.003	0.058
Agency bonds	-	1	0.416	0.127
Corporate bonds	-	-	1	0.003
Equities	-	-	-	1

Table 4b. Correlation matrix for gaps by security type (full sample, 261 observations)

Security type	Treasuries	Agencies	Corporates	Equities
Treasury bonds	1	0.085	0.093	0.013
Agency bonds	-	1	0.126	-0.043
Corporate bonds	-	-	1	0.123
Equities	-	-	-	1

Finally, we regressed the scaled gaps on dummy variables for security type, survey year, and country-specific dummies. Because gaps may also arise from errors in the valuation adjustments we apply, we also include the security-specific price changes applied in each inter-survey period. Results from applying this regression exercise to the sample trimmed of outliers are presented in Table 5. The regression is able to explain only a small fraction of the observed variation in the scaled gaps: the adjusted R $^{2}$ is only .076 and the RMSE comes in at .463: a sizable error for a variable with a mean of .04.¹⁵

Table 5a. Regression of scaled survey gaps on survey, security and country dummy variables, and on price changes: summary statistics

Model: SS	52.064
Model: df	96
Model: MS	0.5423
Residual: SS	361.294
Residual: df	1687
Residual: MS	0.2142
Total: SS	413.358
Total: df	1783
Total: MS	0.2318
Number of obs	1784
F(96, 1687)	2.53
Prob > F	0
R-squared	0.1260
Adj R-squared	0.0762
Root MSE	0.4628

Table 5b. Regression of scaled survey gaps on survey, security and country dummy variables, and on price changes: estimated coefficients

Variable name	Coef.	Std. Err.	t	P>\|t\|	[95% Conf. Interval: Lower]	[95% Conf. Interval: Upper]
constant	0.161	0.107	1.50	0.133	-0.049	0.372
dagcy	-0.084	0.039	-2.16	0.031	-0.160	-0.008
dstk	0.016	0.037	0.43	0.669	-0.057	0.089
dcorp	-0.061	0.038	-1.64	0.102	-0.135	0.012
d1994	-0.014	0.059	-0.24	0.809	-0.130	0.101
d2000	-0.043	0.048	-0.90	0.366	-0.137	0.051
d2002	-0.130	0.048	-2.72	0.007	-0.224	-0.036
d2003	-0.152	0.045	-3.35	0.001	-0.241	-0.063
d2004	-0.039	0.067	-0.58	0.562	-0.170	0.093
d2005	-0.092	0.051	-1.81	0.071	-0.192	0.008
ptreas	0.034	0.573	0.06	0.953	-1.090	1.157
pagcy	-0.538	0.704	-0.76	0.445	-1.919	0.843
pcorp	-0.971	0.530	-1.83	0.067	-2.012	0.069
pstk	-0.074	0.030	-2.49	0.013	-0.133	-0.016
Netherlands	0.228	0.131	1.74	0.082	-0.029	0.484
United Kingdom	-0.421	0.134	-3.15	0.002	-0.683	-0.158
Netherlands Antilles	-0.303	0.143	-2.11	0.035	-0.584	-0.022
Hong Kong	-0.217	0.131	-1.66	0.097	-0.474	0.039
(+ other country dummy variables)	-	-	-	-	-	-

The coefficients for agency and corporate bonds enter with a negative signs (although only the coefficient for agency bonds is significant at the 5 percent level; that for corporate bonds just misses significance at the 10 percent level), providing some statistical confirmation for the observation that on average the estimated positions for agency and corporate bonds tend to overstate the measured survey values for these securities. Of the survey year dummies, we find significant (negative) coefficients for 2002, 2003, and 2005. Most country dummies are not significant; we list a few that are: the U.K. and the Netherlands Antilles, the Netherlands and Hong Kong (although only at the 10 percent level for the last two). Although point estimates for many of the dummy variables seem sensible, confidence bands around them are large. For the U.K. dummy variable, for example, the 95% confidence interval around the point estimate of .421 is -.683 to .158.

Nonetheless, we believe there is some useful information that can be taken away from the regression results. The sizable negative coefficient for the U.K. dummy variable (-.421) indicates that, all else equal, "financial center bias" will lead to an estimated position for the U.K. that overstates the actual survey position by roughly 40 percent. The Netherlands Antilles and Hong Kong also appear to exhibit "financial center bias," while the positive coefficient for the Netherlands indicates that it is a country whose holdings tend to be underestimated. The between-survey price changes for corporate bonds and for equities also enter with significant negative coefficients (only at the 10 percent level for corporate bonds). These results suggest that the prices we impute to foreign holdings of U.S. corporate bonds and U.S. corporate stocks may overstate the actual valuation effects observed for foreign portfolios in these securities: all else equal, price increases will lead to overestimates of the actual survey positions (negative scaled gaps), and price decreases will lead to underestimates. For U.S. stocks, this finding is perhaps not that surprising, given that some of these holdings are actually holdings of mutual funds (including money market mutual funds) and thus may not be appropriately priced by a standard U.S. equity price index. For U.S. corporate bonds, there is no obvious explanation.

4.2 Forecasting the survey gaps

We are left with something of a dilemma: for most countries, there is no clear statistical basis for projecting the survey gaps forward, but for practical purposes we need to make the best forecasts we can. Forecasting the gaps at zero may be appealing on methodological grounds, in the absence of a well-estimated model, but it is clearly unsatisfactory for policy purposes. Even without significant statistical results, we have strong prior beliefs that a large positive or negative gap for, say, the 2002 survey in a given country will not immediately revert to zero in 2003.

We think the best estimates of the gaps will be country and security specific. For example, for U.K. holdings of corporate bonds, we know the gaps are usually large and negative, and we want our forecast of the gap for U.K. bonds holdings to have this property. This suggests using the mean scaled gap for each country and security as the forecast. But as discussed above the gaps are clearly very noisy: there are many outliers, and these can have undue influence on the calculated average gap and thus on the estimated position. In addition, averaging over the entire sample of available surveys ignores more recent developments that may affect the location of transactions (such as the establishment of a new financial center), or that affect reported transactions (such as changes in asset-backed security repayment rates). Such changes may lead to larger or more uniformly positive or negative gaps when observed over the most recent surveys.

In order to address these concerns we adopt another admittedly ad hoc approach: we use the average of the scaled gaps for the most recent liabilities surveys starting with the first annual survey in 2003. This is also the period for which we have asset-backed repayment estimates for agency and corporate bonds, which may introduce a break in the behavior of the gaps. For our in-sample forecasts of 2006 we use the average of three gaps (2003, 2004, and 2005), and going forward from 2006, four gaps.¹⁶

We identify outliers by the same criteria we used to exclude observations from the regression exercise above, but rather than omitting them, we bound both positive and negative gaps at twice the magnitude of the estimated position. While it might be conceptually more appealing to discard these observations entirely, this would in some cases leave us with little data to form an average, and it seems better to try to extract some information from these points. Once the end-of-period gap has been determined (and the assumed end-of-period position has been derived), constructing the monthly positions going forward is straightforward. (By an analogous procedure, one can also extend the estimated positions back, from the earliest survey date to the beginning of the transactions data.)

Figure 8 shows estimated positions based on the June 2005 survey for total foreign holdings of U.S. treasuries and corporate bonds in June 2006 made using the procedure just described (the black line) and for comparison, the "naïve" estimates made using flows and valuation changes only (the red lines). Actual holdings as measured in the June 2006 survey are also shown for comparison.

Figure 8. Forecast of total foreign holdings of U.S. long-term securities, June 2006

Data for Figure 8 immediately follows.

Forecast from June 2005 constructed from adjusted monthly net transactions, valuation adjustments, and moving average of scaled gaps.

Data for Figure 8 - Treasury bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	905	905	-	-	-	905
2002-Jul	-	947	-	-	-	942
2002-Aug	-	967	-	-	-	958
2002-Sep	-	1011	-	-	-	996
2002-Oct	-	1002	-	-	-	982
2002-Nov	-	1009	-	-	-	985
2002-Dec	-	1044	-	-	-	1015
2003-Jan	-	1038	-	-	-	1004
2003-Feb	-	1049	-	-	-	1010
2003-Mar	-	1067	-	-	-	1023
2003-Apr	-	1077	-	-	-	1028
2003-May	-	1145	-	-	-	1090
2003-Jun	1114	1173	1114	-	-	1114
2003-Jul	-	-	1111	-	-	1110
2003-Aug	-	-	1130	-	-	1128
2003-Sep	-	-	1167	-	-	1164
2003-Oct	-	-	1155	-	-	1151
2003-Nov	-	-	1187	-	-	1183
2003-Dec	-	-	1221	-	-	1216
2004-Jan	-	-	1273	-	-	1267
2004-Feb	-	-	1311	-	-	1303
2004-Mar	-	-	1369	-	-	1361
2004-Apr	-	-	1371	-	-	1362
2004-May	-	-	1390	-	-	1380
2004-Jun	1424	-	1435	1424	-	1424
2004-Jul	-	-	-	1445	-	1435
2004-Aug	-	-	-	1483	-	1462
2004-Sep	-	-	-	1499	-	1467
2004-Oct	-	-	-	1525	-	1481
2004-Nov	-	-	-	1532	-	1479
2004-Dec	-	-	-	1550	-	1486
2005-Jan	-	-	-	1589	-	1513
2005-Feb	-	-	-	1615	-	1529
2005-Mar	-	-	-	1638	-	1542
2005-Apr	-	-	-	1691	-	1582
2005-May	-	-	-	1707	-	1585
2005-Jun	1598	-	-	1731	1598	1598
2005-Jul	-	-	-	-	1599	1595
2005-Aug	-	-	-	-	1644	1637
2005-Sep	-	-	-	-	1639	1628
2005-Oct	-	-	-	-	1650	1635
2005-Nov	-	-	-	-	1704	1685
2005-Dec	-	-	-	-	1732	1710
2006-Jan	-	-	-	-	1722	1696
2006-Feb	-	-	-	-	1737	1708
2006-Mar	-	-	-	-	1721	1688
2006-Apr	-	-	-	-	1703	1668
2006-May	-	-	-	-	1712	1674
2006-Jun	1726	-	-	-	1740	1698

Data for Figure 8 - Corporate bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	1130	1130	-	-	-	1130
2002-Jul	-	1129	-	-	-	1117
2002-Aug	-	1162	-	-	-	1138
2002-Sep	-	1179	-	-	-	1142
2002-Oct	-	1169	-	-	-	1120
2002-Nov	-	1193	-	-	-	1132
2002-Dec	-	1230	-	-	-	1155
2003-Jan	-	1250	-	-	-	1163
2003-Feb	-	1276	-	-	-	1175
2003-Mar	-	1294	-	-	-	1181
2003-Apr	-	1327	-	-	-	1200
2003-May	-	1384	-	-	-	1240
2003-Jun	1236	1392	1236	-	-	1236
2003-Jul	-	-	1208	-	-	1214
2003-Aug	-	-	1223	-	-	1235
2003-Sep	-	-	1271	-	-	1290
2003-Oct	-	-	1266	-	-	1291
2003-Nov	-	-	1289	-	-	1320
2003-Dec	-	-	1313	-	-	1351
2004-Jan	-	-	1326	-	-	1371
2004-Feb	-	-	1353	-	-	1405
2004-Mar	-	-	1383	-	-	1441
2004-Apr	-	-	1355	-	-	1417
2004-May	-	-	1357	-	-	1425
2004-Jun	1455	-	1381	1455	-	1455
2004-Jul	-	-	-	1489	-	1485
2004-Aug	-	-	-	1538	-	1529
2004-Sep	-	-	-	1582	-	1568
2004-Oct	-	-	-	1606	-	1587
2004-Nov	-	-	-	1612	-	1589
2004-Dec	-	-	-	1660	-	1632
2005-Jan	-	-	-	1679	-	1646
2005-Feb	-	-	-	1689	-	1652
2005-Mar	-	-	-	1683	-	1641
2005-Apr	-	-	-	1709	-	1662
2005-May	-	-	-	1735	-	1684
2005-Jun	1729	-	-	1786	1729	1729
2005-Jul	-	-	-	-	1720	1718
2005-Aug	-	-	-	-	1765	1760
2005-Sep	-	-	-	-	1771	1763
2005-Oct	-	-	-	-	1774	1763
2005-Nov	-	-	-	-	1805	1792
2005-Dec	-	-	-	-	1842	1825
2006-Jan	-	-	-	-	1854	1835
2006-Feb	-	-	-	-	1883	1861
2006-Mar	-	-	-	-	1896	1871
2006-Apr	-	-	-	-	1915	1887
2006-May	-	-	-	-	1940	1910
2006-Jun	2021	-	-	-	1964	1931

As it happens, the "naïve" estimate derived from applying flows and valuation changes to the 2005 survey value generates a slightly better forecast for June 2006, but such a fortunate outcome cannot be counted on. Furthermore, many individual countries have gaps that are consistently negative or positive, and our procedure takes this somewhat into account. For example, Figure 9 shows that while the forecast positions for the United Kingdom still overestimate the measured survey amounts, they do so by considerably less than the naïve estimates.

Figure 9. Forecast of U.K. holdings of U.S. long-term securities, June 2006

Data for Figure 9 immediately follows.

Forecast from June 2005 constructed from adjusted monthly net transactions, valuation adjustments, and moving average of scaled gaps. Includes Channel Islands and the Isle of Man.

Data for Figure 9 - Treasury bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	48	48	-	-	-	48
2002-Jul	-	61	-	-	-	58
2002-Aug	-	66	-	-	-	61
2002-Sep	-	74	-	-	-	66
2002-Oct	-	78	-	-	-	66
2002-Nov	-	77	-	-	-	62
2002-Dec	-	76	-	-	-	58
2003-Jan	-	77	-	-	-	55
2003-Feb	-	73	-	-	-	49
2003-Mar	-	77	-	-	-	49
2003-Apr	-	74	-	-	-	44
2003-May	-	77	-	-	-	42
2003-Jun	49	86	49	-	-	49
2003-Jul	-	-	69	-	-	64
2003-Aug	-	-	75	-	-	65
2003-Sep	-	-	65	-	-	51
2003-Oct	-	-	69	-	-	50
2003-Nov	-	-	69	-	-	46
2003-Dec	-	-	73	-	-	45
2004-Jan	-	-	83	-	-	50
2004-Feb	-	-	86	-	-	48
2004-Mar	-	-	93	-	-	48
2004-Apr	-	-	105	-	-	55
2004-May	-	-	103	-	-	48
2004-Jun	47	-	109	47	-	47
2004-Jul	-	-	-	52	-	47
2004-Aug	-	-	-	56	-	47
2004-Sep	-	-	-	57	-	43
2004-Oct	-	-	-	61	-	42
2004-Nov	-	-	-	77	-	52
2004-Dec	-	-	-	89	-	57
2005-Jan	-	-	-	88	-	49
2005-Feb	-	-	-	97	-	51
2005-Mar	-	-	-	109	-	55
2005-Apr	-	-	-	113	-	50
2005-May	-	-	-	122	-	49
2005-Jun	53	-	-	137	53	53
2005-Jul	-	-	-	-	67	64
2005-Aug	-	-	-	-	81	75
2005-Sep	-	-	-	-	89	79
2005-Oct	-	-	-	-	93	79
2005-Nov	-	-	-	-	128	108
2005-Dec	-	-	-	-	139	113
2006-Jan	-	-	-	-	149	117
2006-Feb	-	-	-	-	154	115
2006-Mar	-	-	-	-	168	123
2006-Apr	-	-	-	-	154	103
2006-May	-	-	-	-	162	105
2006-Jun	50	-	-	-	188	124

Data for Figure 9 - Corporate bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	108	108	-	-	-	108
2002-Jul	-	108	-	-	-	102
2002-Aug	-	112	-	-	-	101
2002-Sep	-	117	-	-	-	100
2002-Oct	-	119	-	-	-	97
2002-Nov	-	130	-	-	-	102
2002-Dec	-	139	-	-	-	103
2003-Jan	-	149	-	-	-	107
2003-Feb	-	156	-	-	-	106
2003-Mar	-	171	-	-	-	113
2003-Apr	-	186	-	-	-	119
2003-May	-	205	-	-	-	126
2003-Jun	122	209	122	-	-	122
2003-Jul	-	-	126	-	-	125
2003-Aug	-	-	130	-	-	128
2003-Sep	-	-	145	-	-	142
2003-Oct	-	-	149	-	-	145
2003-Nov	-	-	162	-	-	156
2003-Dec	-	-	173	-	-	166
2004-Jan	-	-	173	-	-	165
2004-Feb	-	-	178	-	-	169
2004-Mar	-	-	187	-	-	176
2004-Apr	-	-	189	-	-	177
2004-May	-	-	192	-	-	179
2004-Jun	183	-	197	183	-	183
2004-Jul	-	-	-	194	-	188
2004-Aug	-	-	-	207	-	194
2004-Sep	-	-	-	227	-	207
2004-Oct	-	-	-	237	-	210
2004-Nov	-	-	-	245	-	211
2004-Dec	-	-	-	265	-	223
2005-Jan	-	-	-	270	-	219
2005-Feb	-	-	-	281	-	222
2005-Mar	-	-	-	290	-	223
2005-Apr	-	-	-	300	-	224
2005-May	-	-	-	308	-	223
2005-Jun	241	-	-	336	241	241
2005-Jul	-	-	-	-	245	240
2005-Aug	-	-	-	-	268	259
2005-Sep	-	-	-	-	284	269
2005-Oct	-	-	-	-	294	274
2005-Nov	-	-	-	-	315	291
2005-Dec	-	-	-	-	332	301
2006-Jan	-	-	-	-	343	306
2006-Feb	-	-	-	-	358	316
2006-Mar	-	-	-	-	379	330
2006-Apr	-	-	-	-	394	339
2006-May	-	-	-	-	416	355
2006-Jun	281	-	-	-	433	365

Similarly, as shown in Figure 10, we can significantly reduce the degree of underestimation in holdings of corporate bonds in the euro area.

Figure 10. Forecast of euro area holdings of U.S. long-term securities, June 2006

Data for Figure 10 immediately follows.

Forecast from June 2005 constructed from adjusted monthly net transactions, valuation adjustments, and moving average of scaled gaps.

Data for Figure 10 - Treasury bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	135	135	-	-	-	135
2002-Jul	-	133	-	-	-	133
2002-Aug	-	133	-	-	-	133
2002-Sep	-	137	-	-	-	136
2002-Oct	-	129	-	-	-	129
2002-Nov	-	133	-	-	-	133
2002-Dec	-	141	-	-	-	141
2003-Jan	-	138	-	-	-	138
2003-Feb	-	137	-	-	-	136
2003-Mar	-	132	-	-	-	131
2003-Apr	-	140	-	-	-	139
2003-May	-	146	-	-	-	146
2003-Jun	150	151	150	-	-	150
2003-Jul	-	-	148	-	-	150
2003-Aug	-	-	153	-	-	157
2003-Sep	-	-	155	-	-	160
2003-Oct	-	-	148	-	-	155
2003-Nov	-	-	149	-	-	158
2003-Dec	-	-	140	-	-	150
2004-Jan	-	-	143	-	-	155
2004-Feb	-	-	137	-	-	150
2004-Mar	-	-	137	-	-	152
2004-Apr	-	-	140	-	-	155
2004-May	-	-	138	-	-	155
2004-Jun	154	-	136	154	-	154
2004-Jul	-	-	-	159	-	158
2004-Aug	-	-	-	150	-	148
2004-Sep	-	-	-	151	-	148
2004-Oct	-	-	-	152	-	148
2004-Nov	-	-	-	152	-	147
2004-Dec	-	-	-	153	-	147
2005-Jan	-	-	-	159	-	152
2005-Feb	-	-	-	161	-	153
2005-Mar	-	-	-	161	-	152
2005-Apr	-	-	-	166	-	156
2005-May	-	-	-	175	-	164
2005-Jun	156	-	-	168	156	156
2005-Jul	-	-	-	-	158	158
2005-Aug	-	-	-	-	161	161
2005-Sep	-	-	-	-	161	161
2005-Oct	-	-	-	-	164	165
2005-Nov	-	-	-	-	164	164
2005-Dec	-	-	-	-	168	169
2006-Jan	-	-	-	-	166	167
2006-Feb	-	-	-	-	163	164
2006-Mar	-	-	-	-	158	159
2006-Apr	-	-	-	-	156	157
2006-May	-	-	-	-	157	159
2006-Jun	170	-	-	-	159	160

Data for Figure 10 - Corporate bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	245	245	-	-	-	245
2002-Jul	-	245	-	-	-	252
2002-Aug	-	249	-	-	-	264
2002-Sep	-	252	-	-	-	274
2002-Oct	-	248	-	-	-	277
2002-Nov	-	250	-	-	-	286
2002-Dec	-	256	-	-	-	300
2003-Jan	-	257	-	-	-	309
2003-Feb	-	262	-	-	-	321
2003-Mar	-	262	-	-	-	329
2003-Apr	-	265	-	-	-	339
2003-May	-	273	-	-	-	356
2003-Jun	363	273	363	-	-	363
2003-Jul	-	-	350	-	-	366
2003-Aug	-	-	351	-	-	382
2003-Sep	-	-	361	-	-	409
2003-Oct	-	-	357	-	-	419
2003-Nov	-	-	357	-	-	434
2003-Dec	-	-	362	-	-	454
2004-Jan	-	-	366	-	-	474
2004-Feb	-	-	371	-	-	495
2004-Mar	-	-	377	-	-	517
2004-Apr	-	-	365	-	-	516
2004-May	-	-	363	-	-	526
2004-Jun	545	-	366	545	-	545
2004-Jul	-	-	-	552	-	557
2004-Aug	-	-	-	565	-	576
2004-Sep	-	-	-	572	-	589
2004-Oct	-	-	-	576	-	598
2004-Nov	-	-	-	571	-	598
2004-Dec	-	-	-	578	-	610
2005-Jan	-	-	-	582	-	620
2005-Feb	-	-	-	579	-	622
2005-Mar	-	-	-	572	-	619
2005-Apr	-	-	-	578	-	631
2005-May	-	-	-	584	-	642
2005-Jun	651	-	-	588	651	651
2005-Jul	-	-	-	-	644	652
2005-Aug	-	-	-	-	650	668
2005-Sep	-	-	-	-	643	668
2005-Oct	-	-	-	-	638	672
2005-Nov	-	-	-	-	641	683
2005-Dec	-	-	-	-	647	698
2006-Jan	-	-	-	-	644	704
2006-Feb	-	-	-	-	644	711
2006-Mar	-	-	-	-	635	709
2006-Apr	-	-	-	-	632	713
2006-May	-	-	-	-	630	719
2006-Jun	745	-	-	-	629	726

4.3 Estimating positions for foreign official investors

The same approach to generating between-survey position estimates and forecasts by country can be applied to generate position estimates for holdings of U.S. securities by foreign official institutions. Official investors hold an increasing share of asset-backed securities (especially ABS agency securities), and thus their net transactions should be subject to the same ABS repayment flows.¹⁷ On the other hand, measured official holdings in a new survey typically exceed those expected from summing transactions since the previous survey, even after accounting for valuation changes. This result often occurs because chains of intermediaries can obscure both the type and the country of the foreign holder in the reported transactions data.¹⁸ Figure 11 shows the survey-consistent measures of foreign official holdings of each of the four types of U.S. securities as well as the forecasted value for June 2006. The effect of official net purchases through foreign private intermediaries is especially apparent for official holdings of Treasury and agency bonds since 2002. Although the forecasted values still understate the measured official holdings in June 2006, they do so by less than the "naïve" estimates.

Figure 11. Forecast of holdings of U.S. long-term securities by foreign officials, June 2006

Data for Figure 11 immediately follows.

Forecast from June 2005 constructed from adjusted monthly net transactions, valuation adjustments, and moving average of scaled gaps.

Data for Figure 11 - Treasury bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	557	557	-	-	-	557
2002-Jul	-	562	-	-	-	563
2002-Aug	-	572	-	-	-	574
2002-Sep	-	581	-	-	-	585
2002-Oct	-	573	-	-	-	577
2002-Nov	-	577	-	-	-	582
2002-Dec	-	596	-	-	-	602
2003-Jan	-	594	-	-	-	601
2003-Feb	-	606	-	-	-	614
2003-Mar	-	601	-	-	-	610
2003-Apr	-	600	-	-	-	610
2003-May	-	629	-	-	-	640
2003-Jun	650	638	650	-	-	650
2003-Jul	-	-	631	-	-	640
2003-Aug	-	-	628	-	-	645
2003-Sep	-	-	655	-	-	681
2003-Oct	-	-	666	-	-	701
2003-Nov	-	-	683	-	-	727
2003-Dec	-	-	695	-	-	749
2004-Jan	-	-	725	-	-	788
2004-Feb	-	-	750	-	-	823
2004-Mar	-	-	788	-	-	871
2004-Apr	-	-	782	-	-	873
2004-May	-	-	784	-	-	884
2004-Jun	910	-	801	910	-	910
2004-Jul	-	-	-	922	-	924
2004-Aug	-	-	-	953	-	957
2004-Sep	-	-	-	963	-	969
2004-Oct	-	-	-	983	-	990
2004-Nov	-	-	-	987	-	997
2004-Dec	-	-	-	1000	-	1012
2005-Jan	-	-	-	1011	-	1025
2005-Feb	-	-	-	1012	-	1027
2005-Mar	-	-	-	991	-	1008
2005-Apr	-	-	-	1018	-	1037
2005-May	-	-	-	1034	-	1056
2005-Jun	1078	-	-	1055	1078	1078
2005-Jul	-	-	-	-	1064	1067
2005-Aug	-	-	-	-	1081	1087
2005-Sep	-	-	-	-	1062	1071
2005-Oct	-	-	-	-	1056	1067
2005-Nov	-	-	-	-	1062	1076
2005-Dec	-	-	-	-	1074	1092
2006-Jan	-	-	-	-	1074	1093
2006-Feb	-	-	-	-	1085	1108
2006-Mar	-	-	-	-	1062	1087
2006-Apr	-	-	-	-	1066	1094
2006-May	-	-	-	-	1050	1080
2006-Jun	1212	-	-	-	1046	1079

Data for Figure 11 - Agency bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	134	134	-	-	-	134
2002-Jul	-	138	-	-	-	140
2002-Aug	-	144	-	-	-	146
2002-Sep	-	149	-	-	-	152
2002-Oct	-	149	-	-	-	153
2002-Nov	-	148	-	-	-	153
2002-Dec	-	153	-	-	-	160
2003-Jan	-	157	-	-	-	165
2003-Feb	-	163	-	-	-	172
2003-Mar	-	163	-	-	-	173
2003-Apr	-	164	-	-	-	176
2003-May	-	167	-	-	-	179
2003-Jun	180	166	180	-	-	180
2003-Jul	-	-	173	-	-	175
2003-Aug	-	-	174	-	-	178
2003-Sep	-	-	179	-	-	185
2003-Oct	-	-	179	-	-	187
2003-Nov	-	-	180	-	-	190
2003-Dec	-	-	185	-	-	197
2004-Jan	-	-	189	-	-	203
2004-Feb	-	-	194	-	-	210
2004-Mar	-	-	197	-	-	215
2004-Apr	-	-	193	-	-	213
2004-May	-	-	189	-	-	210
2004-Jun	211	-	188	211	-	211
2004-Jul	-	-	-	214	-	222
2004-Aug	-	-	-	219	-	235
2004-Sep	-	-	-	220	-	244
2004-Oct	-	-	-	220	-	251
2004-Nov	-	-	-	220	-	259
2004-Dec	-	-	-	221	-	267
2005-Jan	-	-	-	227	-	281
2005-Feb	-	-	-	228	-	289
2005-Mar	-	-	-	227	-	295
2005-Apr	-	-	-	226	-	303
2005-May	-	-	-	230	-	315
2005-Jun	324	-	-	232	324	324
2005-Jul	-	-	-	-	323	327
2005-Aug	-	-	-	-	323	331
2005-Sep	-	-	-	-	320	331
2005-Oct	-	-	-	-	318	333
2005-Nov	-	-	-	-	320	338
2005-Dec	-	-	-	-	322	344
2006-Jan	-	-	-	-	329	354
2006-Feb	-	-	-	-	331	359
2006-Mar	-	-	-	-	330	362
2006-Apr	-	-	-	-	333	369
2006-May	-	-	-	-	339	378
2006-Jun	473	-	-	-	343	385

Data for Figure 11 - Corporate bonds

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	18	18	-	-	-	18
2002-Jul	-	19	-	-	-	18
2002-Aug	-	20	-	-	-	19
2002-Sep	-	20	-	-	-	19
2002-Oct	-	20	-	-	-	19
2002-Nov	-	21	-	-	-	20
2002-Dec	-	21	-	-	-	20
2003-Jan	-	22	-	-	-	20
2003-Feb	-	22	-	-	-	21
2003-Mar	-	23	-	-	-	21
2003-Apr	-	23	-	-	-	21
2003-May	-	24	-	-	-	21
2003-Jun	21	24	21	-	-	21
2003-Jul	-	-	21	-	-	22
2003-Aug	-	-	21	-	-	23
2003-Sep	-	-	22	-	-	25
2003-Oct	-	-	22	-	-	27
2003-Nov	-	-	23	-	-	29
2003-Dec	-	-	24	-	-	31
2004-Jan	-	-	25	-	-	33
2004-Feb	-	-	25	-	-	35
2004-Mar	-	-	27	-	-	37
2004-Apr	-	-	26	-	-	38
2004-May	-	-	27	-	-	39
2004-Jun	41	-	27	41	-	41
2004-Jul	-	-	-	43	-	43
2004-Aug	-	-	-	44	-	45
2004-Sep	-	-	-	46	-	46
2004-Oct	-	-	-	47	-	48
2004-Nov	-	-	-	48	-	49
2004-Dec	-	-	-	50	-	52
2005-Jan	-	-	-	52	-	53
2005-Feb	-	-	-	53	-	55
2005-Mar	-	-	-	52	-	54
2005-Apr	-	-	-	53	-	55
2005-May	-	-	-	55	-	58
2005-Jun	61	-	-	58	61	61
2005-Jul	-	-	-	-	61	62
2005-Aug	-	-	-	-	64	65
2005-Sep	-	-	-	-	65	67
2005-Oct	-	-	-	-	66	68
2005-Nov	-	-	-	-	67	70
2005-Dec	-	-	-	-	70	74
2006-Jan	-	-	-	-	72	76
2006-Feb	-	-	-	-	76	81
2006-Mar	-	-	-	-	77	83
2006-Apr	-	-	-	-	79	84
2006-May	-	-	-	-	81	87
2006-Jun	96	-	-	-	82	89

Data for Figure 11 - Equities

Date	Survey value	Survey date 2002-Jun: net flows (adj.)	Survey date 2003-Jun: net flows (adj.)	Survey date 2004-Jun: net flows (adj.)	Survey date 2005-Jun: net flows (adj.)	Survey date 2002-Jun: estimated position
2002-Jun	84	84	-	-	-	84
2002-Jul	-	78	-	-	-	80
2002-Aug	-	78	-	-	-	82
2002-Sep	-	69	-	-	-	74
2002-Oct	-	75	-	-	-	82
2002-Nov	-	80	-	-	-	89
2002-Dec	-	75	-	-	-	85
2003-Jan	-	73	-	-	-	84
2003-Feb	-	72	-	-	-	85
2003-Mar	-	72	-	-	-	87
2003-Apr	-	78	-	-	-	96
2003-May	-	82	-	-	-	102
2003-Jun	105	83	105	-	-	105
2003-Jul	-	-	107	-	-	108
2003-Aug	-	-	109	-	-	111
2003-Sep	-	-	108	-	-	110
2003-Oct	-	-	113	-	-	116
2003-Nov	-	-	114	-	-	118
2003-Dec	-	-	120	-	-	124
2004-Jan	-	-	121	-	-	127
2004-Feb	-	-	123	-	-	129
2004-Mar	-	-	121	-	-	128
2004-Apr	-	-	120	-	-	127
2004-May	-	-	121	-	-	130
2004-Jun	132	-	123	132	-	132
2004-Jul	-	-	-	128	-	131
2004-Aug	-	-	-	128	-	134
2004-Sep	-	-	-	130	-	139
2004-Oct	-	-	-	131	-	143
2004-Nov	-	-	-	137	-	153
2004-Dec	-	-	-	142	-	162
2005-Jan	-	-	-	138	-	160
2005-Feb	-	-	-	141	-	166
2005-Mar	-	-	-	139	-	167
2005-Apr	-	-	-	135	-	165
2005-May	-	-	-	139	-	174
2005-Jun	177	-	-	140	177	177
2005-Jul	-	-	-	-	184	185
2005-Aug	-	-	-	-	182	185
2005-Sep	-	-	-	-	184	189
2005-Oct	-	-	-	-	181	188
2005-Nov	-	-	-	-	188	197
2005-Dec	-	-	-	-	188	198
2006-Jan	-	-	-	-	194	206
2006-Feb	-	-	-	-	193	207
2006-Mar	-	-	-	-	197	213
2006-Apr	-	-	-	-	203	220
2006-May	-	-	-	-	197	216
2006-Jun	215	-	-	-	198	219

5 Estimating positions on the U.S. claims side

Figure 12 presents estimated positions using the same basic methodology for U.S. claims on foreigners. As we discuss in Appendix B, we allow for country-specific valuation adjustments in constructing our claims estimates and then because our methodology adjusts for financial center (transactions) bias, our estimated holdings will provide a reasonable time series approximation of actual holdings by country of issuer. One difference from our liabilities estimates is that we do construct the aggregate U.S. portfolio of foreign securities by summing the estimated holdings of securities of each individual country.¹⁹ The advantage to this approach on the claims side is that that because we allow for country-specific valuation effects, the resulting aggregate portfolio is weighted according to the actual countries of issue of the securities that U.S. investors hold. The resulting valuation changes we impute will better approximate the actual price and exchange rate movements realized by U.S. investors than will most global bond or equity indexes: for example, our constructed foreign bond portfolio is roughly 70 percent dollar-denominated, compared with approximately 25 percent for a standard global bond index such as the MSCI World Sovereign bond index. Our approach to estimating positions for U.S. claims after December 2005 is also similar to that for U.S. liabilities, but because claims surveys have only been conducted annually beginning in 2004, the average gap is constructed using only the gaps for 2004 and 2005. Figure 13 shows similar results for U.S. claims on the United Kingdom.

Figure 12. Forecast of total U.S. holdings of foreign long-term securities, December 2006

Data for Figure 12 immediately follows.

Forecast from December 2005 constructed from adjusted monthly net transactions, valuation adjustments, and moving average of scaled gaps.

Data for Figure 12 - Bonds

Date	Survey value	Survey date 2001-Dec: net flows (adj.)	Survey date 2003-Dec: net flows (adj.)	Survey date 2004-Dec: net flows (adj.)	Survey date 2005-Dec: net flows (adj.)	Survey date 2001-Dec: estimated position
2001-Dec	557	557	-	-	-	557
2002-Jan	-	559	-	-	-	571
2002-Feb	-	558	-	-	-	583
2002-Mar	-	543	-	-	-	578
2002-Apr	-	554	-	-	-	600
2002-May	-	553	-	-	-	610
2002-Jun	-	559	-	-	-	629
2002-Jul	-	552	-	-	-	633
2002-Aug	-	565	-	-	-	657
2002-Sep	-	569	-	-	-	673
2002-Oct	-	561	-	-	-	674
2002-Nov	-	559	-	-	-	682
2002-Dec	-	587	-	-	-	724
2003-Jan	-	591	-	-	-	740
2003-Feb	-	583	-	-	-	742
2003-Mar	-	583	-	-	-	752
2003-Apr	-	590	-	-	-	771
2003-May	-	610	-	-	-	808
2003-Jun	-	610	-	-	-	816
2003-Jul	-	574	-	-	-	780
2003-Aug	-	568	-	-	-	781
2003-Sep	-	581	-	-	-	815
2003-Oct	-	586	-	-	-	825
2003-Nov	-	598	-	-	-	847
2003-Dec	874	610	874	-	-	874
2004-Jan	-	-	879	-	-	886
2004-Feb	-	-	895	-	-	903
2004-Mar	-	-	898	-	-	909
2004-Apr	-	-	853	-	-	866
2004-May	-	-	855	-	-	868
2004-Jun	-	-	855	-	-	869
2004-Jul	-	-	868	-	-	882
2004-Aug	-	-	891	-	-	906
2004-Sep	-	-	925	-	-	939
2004-Oct	-	-	946	-	-	957
2004-Nov	-	-	971	-	-	977
2004-Dec	993	-	990	993	-	993
2005-Jan	-	-	-	983	-	989
2005-Feb	-	-	-	984	-	994
2005-Mar	-	-	-	973	-	988
2005-Apr	-	-	-	991	-	1010
2005-May	-	-	-	993	-	1016
2005-Jun	-	-	-	997	-	1024
2005-Jul	-	-	-	993	-	1022
2005-Aug	-	-	-	993	-	1026
2005-Sep	-	-	-	997	-	1031
2005-Oct	-	-	-	980	-	1017
2005-Nov	-	-	-	976	-	1014
2005-Dec	1028	-	-	987	1028	1028
2006-Jan	-	-	-	-	1042	1052
2006-Feb	-	-	-	-	1038	1055
2006-Mar	-	-	-	-	1027	1052
2006-Apr	-	-	-	-	1044	1077
2006-May	-	-	-	-	1069	1109
2006-Jun	-	-	-	-	1068	1116
2006-Jul	-	-	-	-	1098	1153
2006-Aug	-	-	-	-	1126	1191
2006-Sep	-	-	-	-	1137	1212
2006-Oct	-	-	-	-	1158	1242
2006-Nov	-	-	-	-	1189	1281
2006-Dec	1294	-	-	-	1204	1307

Data for Figure 12 - Equities

Date	Survey value	Survey date 2001-Dec: net flows (adj.)	Survey date 2003-Dec: net flows (adj.)	Survey date 2004-Dec: net flows (adj.)	Survey date 2005-Dec: net flows (adj.)	Survey date 2001-Dec: estimated position
2001-Dec	1613	1613	-	-	-	1613
2002-Jan	-	1503	-	-	-	1513
2002-Feb	-	1499	-	-	-	1518
2002-Mar	-	1578	-	-	-	1609
2002-Apr	-	1550	-	-	-	1584
2002-May	-	1566	-	-	-	1611
2002-Jun	-	1477	-	-	-	1522
2002-Jul	-	1318	-	-	-	1366
2002-Aug	-	1317	-	-	-	1372
2002-Sep	-	1175	-	-	-	1230
2002-Oct	-	1253	-	-	-	1316
2002-Nov	-	1320	-	-	-	1395
2002-Dec	-	1272	-	-	-	1350
2003-Jan	-	1230	-	-	-	1311
2003-Feb	-	1207	-	-	-	1294
2003-Mar	-	1200	-	-	-	1288
2003-Apr	-	1313	-	-	-	1416
2003-May	-	1411	-	-	-	1530
2003-Jun	-	1448	-	-	-	1579
2003-Jul	-	1487	-	-	-	1628
2003-Aug	-	1542	-	-	-	1696
2003-Sep	-	1589	-	-	-	1752
2003-Oct	-	1697	-	-	-	1875
2003-Nov	-	1732	-	-	-	1921
2003-Dec	2079	1869	2079	-	-	2079
2004-Jan	-	-	2072	-	-	2077
2004-Feb	-	-	2134	-	-	2143
2004-Mar	-	-	2125	-	-	2139
2004-Apr	-	-	2065	-	-	2084
2004-May	-	-	2071	-	-	2095
2004-Jun	-	-	2127	-	-	2158
2004-Jul	-	-	2075	-	-	2108
2004-Aug	-	-	2087	-	-	2124
2004-Sep	-	-	2168	-	-	2211
2004-Oct	-	-	2245	-	-	2293
2004-Nov	-	-	2380	-	-	2439
2004-Dec	2560	-	2492	2560	-	2560
2005-Jan	-	-	-	2527	-	2544
2005-Feb	-	-	-	2662	-	2698
2005-Mar	-	-	-	2598	-	2652
2005-Apr	-	-	-	2517	-	2585
2005-May	-	-	-	2541	-	2627
2005-Jun	-	-	-	2600	-	2704
2005-Jul	-	-	-	2712	-	2838
2005-Aug	-	-	-	2774	-	2919
2005-Sep	-	-	-	2906	-	3074
2005-Oct	-	-	-	2809	-	2990
2005-Nov	-	-	-	2940	-	3147
2005-Dec	3318	-	-	3084	3318	3318
2006-Jan	-	-	-	-	3568	3584
2006-Feb	-	-	-	-	3555	3587
2006-Mar	-	-	-	-	3653	3701
2006-Apr	-	-	-	-	3834	3900
2006-May	-	-	-	-	3644	3723
2006-Jun	-	-	-	-	3632	3725
2006-Jul	-	-	-	-	3661	3767
2006-Aug	-	-	-	-	3735	3856
2006-Sep	-	-	-	-	3761	3900
2006-Oct	-	-	-	-	3913	4071
2006-Nov	-	-	-	-	4089	4271
2006-Dec	4329	-	-	-	4192	4390

Figure 13. Forecast of total U.S. holdings of U.K. long-term securities, December 2006

Data for Figure 13 immediately follows.

Forecast from December 2005 constructed from adjusted monthly net transactions, valuation adjustments, and moving average of scaled gaps. Includes Channel Islands and the Isle of Man.

Data for Figure 13 - Bonds

Date	Survey value	Survey date 2001-Dec: net flows (adj.)	Survey date 2003-Dec: net flows (adj.)	Survey date 2004-Dec: net flows (adj.)	Survey date 2005-Dec: net flows (adj.)	Survey date 2001-Dec: estimated position
2001-Dec	83	83	-	-	-	83
2002-Jan	-	83	-	-	-	87
2002-Feb	-	83	-	-	-	90
2002-Mar	-	76	-	-	-	86
2002-Apr	-	79	-	-	-	93
2002-May	-	77	-	-	-	93
2002-Jun	-	80	-	-	-	100
2002-Jul	-	80	-	-	-	105
2002-Aug	-	83	-	-	-	111
2002-Sep	-	83	-	-	-	116
2002-Oct	-	80	-	-	-	115
2002-Nov	-	74	-	-	-	111
2002-Dec	-	81	-	-	-	123
2003-Jan	-	76	-	-	-	122
2003-Feb	-	65	-	-	-	112
2003-Mar	-	66	-	-	-	115
2003-Apr	-	66	-	-	-	117
2003-May	-	70	-	-	-	126
2003-Jun	-	74	-	-	-	133
2003-Jul	-	63	-	-	-	121
2003-Aug	-	64	-	-	-	123
2003-Sep	-	62	-	-	-	127
2003-Oct	-	62	-	-	-	128
2003-Nov	-	67	-	-	-	136
2003-Dec	148	73	148	-	-	148
2004-Jan	-	-	153	-	-	149
2004-Feb	-	-	174	-	-	165
2004-Mar	-	-	168	-	-	154
2004-Apr	-	-	160	-	-	143
2004-May	-	-	166	-	-	144
2004-Jun	-	-	168	-	-	142
2004-Jul	-	-	178	-	-	147
2004-Aug	-	-	191	-	-	155
2004-Sep	-	-	212	-	-	170
2004-Oct	-	-	222	-	-	174
2004-Nov	-	-	237	-	-	180
2004-Dec	179	-	242	179	-	179
2005-Jan	-	-	-	177	-	178
2005-Feb	-	-	-	180	-	183
2005-Mar	-	-	-	182	-	185
2005-Apr	-	-	-	193	-	197
2005-May	-	-	-	194	-	200
2005-Jun	-	-	-	201	-	208
2005-Jul	-	-	-	196	-	204
2005-Aug	-	-	-	189	-	199
2005-Sep	-	-	-	186	-	197
2005-Oct	-	-	-	180	-	192
2005-Nov	-	-	-	176	-	188
2005-Dec	194	-	-	181	194	194
2006-Jan	-	-	-	-	201	200
2006-Feb	-	-	-	-	200	197
2006-Mar	-	-	-	-	205	201
2006-Apr	-	-	-	-	212	207
2006-May	-	-	-	-	234	227
2006-Jun	-	-	-	-	235	227
2006-Jul	-	-	-	-	251	241
2006-Aug	-	-	-	-	270	257
2006-Sep	-	-	-	-	273	259
2006-Oct	-	-	-	-	281	266
2006-Nov	-	-	-	-	302	284
2006-Dec	256	-	-	-	312	293

Data for Figure 13 - Equities

Date	Survey value	Survey date 2001-Dec: net flows (adj.)	Survey date 2003-Dec: net flows (adj.)	Survey date 2004-Dec: net flows (adj.)	Survey date 2005-Dec: net flows (adj.)	Survey date 2001-Dec: estimated position
2001-Dec	355	355	-	-	-	355
2002-Jan	-	341	-	-	-	341
2002-Feb	-	341	-	-	-	341
2002-Mar	-	349	-	-	-	349
2002-Apr	-	355	-	-	-	355
2002-May	-	352	-	-	-	352
2002-Jun	-	340	-	-	-	340
2002-Jul	-	309	-	-	-	309
2002-Aug	-	303	-	-	-	303
2002-Sep	-	272	-	-	-	272
2002-Oct	-	297	-	-	-	297
2002-Nov	-	305	-	-	-	305
2002-Dec	-	299	-	-	-	300
2003-Jan	-	283	-	-	-	283
2003-Feb	-	278	-	-	-	278
2003-Mar	-	291	-	-	-	291
2003-Apr	-	318	-	-	-	318
2003-May	-	341	-	-	-	342
2003-Jun	-	346	-	-	-	347
2003-Jul	-	354	-	-	-	354
2003-Aug	-	352	-	-	-	352
2003-Sep	-	360	-	-	-	361
2003-Oct	-	388	-	-	-	388
2003-Nov	-	395	-	-	-	395
2003-Dec	426	426	426	-	-	426
2004-Jan	-	-	429	-	-	424
2004-Feb	-	-	449	-	-	438
2004-Mar	-	-	427	-	-	411
2004-Apr	-	-	428	-	-	407
2004-May	-	-	440	-	-	413
2004-Jun	-	-	447	-	-	415
2004-Jul	-	-	447	-	-	411
2004-Aug	-	-	449	-	-	408
2004-Sep	-	-	463	-	-	415
2004-Oct	-	-	479	-	-	425
2004-Nov	-	-	510	-	-	447
2004-Dec	462	-	533	462	-	462
2005-Jan	-	-	-	459	-	462
2005-Feb	-	-	-	488	-	494
2005-Mar	-	-	-	478	-	487
2005-Apr	-	-	-	471	-	483
2005-May	-	-	-	465	-	480
2005-Jun	-	-	-	476	-	494
2005-Jul	-	-	-	482	-	504
2005-Aug	-	-	-	498	-	523
2005-Sep	-	-	-	507	-	535
2005-Oct	-	-	-	488	-	519
2005-Nov	-	-	-	489	-	522
2005-Dec	545	-	-	507	545	545
2006-Jan	-	-	-	-	579	578
2006-Feb	-	-	-	-	576	574
2006-Mar	-	-	-	-	591	587
2006-Apr	-	-	-	-	627	622
2006-May	-	-	-	-	616	611
2006-Jun	-	-	-	-	627	620
2006-Jul	-	-	-	-	644	636
2006-Aug	-	-	-	-	654	645
2006-Sep	-	-	-	-	659	649
2006-Oct	-	-	-	-	698	686
2006-Nov	-	-	-	-	722	709
2006-Dec	689	-	-	-	750	735

6 Decomposing changes between surveys into flows, valuation changes, and other factors

The methodology we have described provides a straightforward way to construct consistent monthly between-survey estimates of securities positions, but it leaves open the question of how to decompose the evolution of the change in position into flows, valuation effects, and other factors: an exercise of considerable interest to analysts attempting to reconcile U.S. cross-border financial flows with the evolution of the cross border holdings such as that presented in the U.S. international investment position (IIP).²⁰ We identify "flows" as the adjusted net transactions data that include not only the appropriate adjustments for stock swaps and ABS repayments, but also the corrections we make to the flow data to prevent the between-survey positions from becoming negative.²¹ We can also get a measure of the size and direction of the valuation effect over the period in question from the difference between the valuation-adjusted position estimate and a flow-only estimate, such as that illustrated in Figure 2. Of course, accounting for these identifiable factors still leaves the observed gap at the end of the period. But because our methodology distributes the gap month by month, we can decompose the change between our estimated positions for any two periods into the contributions from flows, valuation changes, and the gap. Thus, although our liabilities surveys initially were conducted only once every five years, and more recently are conducted annually but as of end-June, we can decompose the annual changes in our year-end position estimates into flows, valuation effects, and the "gap" similar to the presentation in the IIP. Figures 14 and 15 summarize these changes for total foreign holdings and U.K. holdings of each type of U.S. long-term security.

For total foreign holdings of U.S. debt securities (Figure 14a-c), the estimated change in position in most years is primarily accounted for by financial flows, whereas valuation effects account for the bulk of the change in foreign holdings of U.S. equities, especially in recent years (Figure 14d). The charts also indicate that at times the "gap" can make a sizable contribution. Because the gap itself may arise from flow errors or valuation errors, it may be desirable to attempt to decompose the gap error into the relevant contributing factors. For this purpose, the regression results from Table 3 may provide some guidance, but unfortunately the limited ability of the regression exercise to explain much of the variation and the imprecision with which the coefficients are estimated makes it difficult to draw firm conclusions.

Figure 14a. Foreign holdings of U.S. treasury bonds
Components of change in year-end position in billions of dollars

Data for Figure 14a immediately follows.

Data for Figure 14a

Date	Total change	Net flows	Valuation changes	Estimated gap
1985	46	29	12	5
1986	35	20	9	7
1987	17	26	-12	3
1988	47	49	-5	3
1989	70	54	14	1
1990	5	19	-1	-13
1991	29	21	23	-16
1992	30	48	-1	-17
1993	32	29	16	-14
1994	36	103	-50	-17
1995	166	136	55	-25
1996	178	234	-25	-32
1997	164	189	29	-53
1998	27	53	38	-63
1999	-134	-8	-98	-28
2000	-13	-53	59	-19
2001	19	25	4	-9
2002	150	149	55	-54
2003	214	277	-26	-37
2004	263	369	-12	-93
2005	240	353	-23	-89
2006	169	229	-29	-32

Figure 14b. Foreign holdings of U.S. agency bonds
Components of change in year-end position in billions of dollars

Data for Figure 14b immediately follows.

Data for Figure 14b

Date	Total change	Net flows	Valuation changes	Estimated gap
1985	4	4	1	-1
1986	7	7	1	-1
1987	3	5	-1	-1
1988	5	7	-1	-1
1989	16	15	2	-0
1990	3	6	1	-4
1991	10	10	4	-5
1992	13	19	-0	-5
1993	31	36	3	-7
1994	2	22	-13	-7
1995	20	29	12	-21
1996	15	42	-4	-23
1997	25	50	6	-32
1998	28	58	7	-37
1999	41	92	-23	-28
2000	127	160	19	-53
2001	88	164	10	-86
2002	114	169	25	-80
2003	12	80	-7	-61
2004	152	179	-4	-22
2005	154	156	-16	14
2006	240	235	-2	7

Figure 14c. Foreign holdings of U.S. corporate bonds
Components of change in year-end position in billions of dollars

Data for Figure 14c immediately follows.

Data for Figure 14c

Date	Total change	Net flows	Valuation changes	Estimated gap
1985	61	40	6	15
1986	50	43	5	1
1987	13	22	-9	0
1988	18	21	-1	-2
1989	17	18	7	-8
1990	11	10	-2	2
1991	20	18	7	-5
1992	13	21	0	-8
1993	26	33	5	-13
1994	14	42	-20	-8
1995	82	58	20	4
1996	69	84	-7	-8
1997	68	84	9	-26
1998	95	123	8	-36
1999	87	161	-48	-26
2000	185	198	9	-22
2001	193	222	17	-46
2002	106	168	22	-84
2003	195	221	11	-36
2004	282	272	5	5
2005	247	292	-47	3
2006	463	423	-27	67

Figure 14d. Foreign holdings of U.S. equities
Components of change in year-end position in billions of dollars

Data for Figure 14d immediately follows.

Data for Figure 14d

Date	Total change	Net flows	Valuation changes	Estimated gap
1985	33	5	29	-0
1986	37	19	19	-1
1987	13	17	-4	-1
1988	20	-2	22	-1
1989	68	11	57	-0
1990	-28	-15	-16	3
1991	84	12	68	4
1992	12	-4	13	3
1993	50	22	24	3
1994	4	4	-4	3
1995	156	25	138	-7
1996	136	28	121	-14
1997	285	79	232	-27
1998	317	58	300	-41
1999	365	116	298	-50
2000	-24	193	-248	31
2001	-55	124	-210	31
2002	-245	57	-368	67
2003	507	37	358	112
2004	287	65	166	56
2005	183	86	85	12
2006	551	155	317	80

Figure 15a. U.K. holdings of U.S. treasury bonds
Components of change in year-end position in billions of dollars

Data for Figure 15a immediately follows.

Includes Channel Islands and the Isle of Man.

Data for Figure 15a

Date	Total change	Net flows	Valuation changes	Estimated gap
1985	-4	-2	1	-3
1986	2	5	0	-3
1987	0	4	-1	-3
1988	5	10	-0	-5
1989	14	20	1	-8
1990	-7	-2	-0	-5
1991	2	6	2	-5
1992	18	24	0	-7
1993	-2	6	2	-10
1994	12	23	-6	-5
1995	24	35	7	-18
1996	41	66	-3	-22
1997	67	98	7	-38
1998	-23	14	9	-47
1999	-67	-20	-23	-23
2000	-51	-34	7	-25
2001	-20	-7	-0	-12
2002	40	62	4	-26
2003	-13	34	-0	-46
2004	12	80	-2	-66
2005	27	135	-0	-108
2006	-27	93	-5	-114

Figure 15b. U.K. holdings of U.S. agency bonds
Components of change in year-end position in billions of dollars

Data for Figure 15b immediately follows.

Includes Channel Islands and the Isle of Man.

Data for Figure 15b

Date	Total change	Net flows	Valuation changes	Estimated gap
1985	1	2	0	-1
1986	-1	1	0	-2
1987	-1	1	-0	-2
1988	-1	1	-0	-2
1989	3	5	0	-3
1990	0	2	0	-2
1991	-0	1	0	-2
1992	2	4	0	-2
1993	4	7	0	-3
1994	1	5	-2	-3
1995	12	17	2	-7
1996	2	14	-1	-11
1997	-3	10	2	-14
1998	17	32	2	-17
1999	-7	14	-6	-15
2000	10	30	3	-23
2001	6	44	1	-40
2002	-12	30	3	-45
2003	-18	20	0	-38
2004	20	51	-0	-30
2005	-8	37	-1	-45
2006	4	50	-1	-45

Figure 15c. U.K. holdings of U.S. corporate bonds
Components of change in year-end position in billions of dollars

Data for Figure 15c immediately follows.

Includes Channel Islands and the Isle of Man.

Data for Figure 15c

Date	Total change	Net flows	Valuation changes	Estimated gap
1985	25	30	2	-8
1986	16	33	3	-20
1987	-9	18	-5	-21
1988	-12	12	-1	-23
1989	-12	12	4	-28
1990	2	8	-0	-6
1991	1	7	2	-9
1992	-1	9	0	-10
1993	1	12	1	-12
1994	10	26	-7	-9
1995	26	40	5	-20
1996	14	47	-2	-31
1997	3	45	4	-45
1998	15	68	4	-57
1999	20	92	-23	-50
2000	35	111	2	-79
2001	-13	110	5	-129
2002	-21	84	2	-107
2003	64	118	4	-58
2004	56	107	-0	-50
2005	40	169	-8	-121
2006	98	248	-9	-141

Figure 15d. U.K. holdings of U.S. equities
Components of change in year-end position in billions of dollars

Data for Figure 15d immediately follows.

Includes Channel Islands and the Isle of Man.

Data for Figure 15d

Date	Total change	Net flows	Valuation changes	Estimated gap
1985	9	2	8	-1
1986	10	5	5	-1
1987	-1	1	-1	-1
1988	4	-1	6	-1
1989	17	4	15	-2
1990	-6	-3	-4	1
1991	18	1	17	1
1992	1	-3	3	1
1993	11	5	5	1
1994	1	1	-1	1
1995	29	9	33	-13
1996	17	6	29	-18
1997	50	23	56	-29
1998	62	28	75	-41
1999	73	44	78	-49
2000	-20	71	-47	-44
2001	-46	38	-47	-37
2002	-67	15	-72	-9
2003	58	1	50	7
2004	43	26	23	-6
2005	-1	20	11	-32
2006	94	76	43	-25

Where we identify significant individual country dummy variables (such as for the U.K.), these regression results indicate systematic but country-specific problems with the flow data: we interpret this result for the U.K. as the financial center bias that leads to systematic overestimates for U.K. holdings. Thus, in Figures 15a-15d for changes in the U.K. position estimates, the large negative gap contributions about offset the positive flow contributions, leaving the total change in position (denoted by the solid line) showing a much smaller annual change. The overall gaps we observe for the total of all countries are a bit more problematic to sort out. The negative coefficients for agency and corporate bonds in the regression exercise correspond to the relatively persistent negative gap contributions for changes in foreign holdings of these securities. These gaps may in part reflect the imprecision with which we can adjust for ABS repayments (and the fact that such ABS repayment flows are not available prior to 2002), and thus these gaps may plausibly offset some of the flow changes for these securities as well. The significant (negative) survey year dummies for 2002, 2003, and 2005 may suggest that there were particular problems with those year's measured survey values, but may also indicate changes in reporting of the flow data. In practice, each year's survey data are used by the compilers of the data to cross-check reporting on the flow data, and when the survey results point to a particular problem with the flow data, reporting instructions may be clarified for data going forward.²² As we noted above, the negative price coefficients for corporate bonds and for stocks suggest that we may over-estimate the valuation changes that apply to these securities, and may in part account for the tendency for the gap contributions to partially offset the valuation contributions for these securities. However, we reiterate that the confidence with which we can assign such contributions to the gaps is limited, and necessarily will be a judgmental exercise.²³

Figure 16 repeats the decomposition exercise for U.S. holdings of foreign securities (because the claims surveys only begin in 1994, we can provide this decomposition only from 1995). A few observations stand out. First, for U.S. holdings of foreign equity,

Figure 16a. U.S. holdings of foreign bonds
Components of change in year-end position in billions of dollars

Data for Figure 16a immediately follows.

Data for Figure 16a

Date	Total change	Net flows	Valuation changes	Estimated gap
1995	107	51	39	16
1996	83	53	18	12
1997	40	50	-16	6
1998	47	21	12	14
1999	-17	10	-30	3
2000	10	12	-5	3
2001	-29	-10	-21	2
2002	167	-18	58	127
2003	150	-11	54	107
2004	119	83	48	-12
2005	35	55	-51	31
2006	266	162	8	96

valuation effects make a very sizable contribution, reflecting both the effects of foreign equity price changes (in home currencies) as well as effects of the exchange value of the dollar, with the appreciation of the dollar contributing to the negative valuation effects in 2000 and 2001, and dollar depreciation mitigating the negative price contribution in 2002 and contributing to the positive valuation effects in 2003 and 2004. Second, the gap

Figure 16b. U.S. holdings of foreign equities
Components of change in year-end position in billions of dollars

Data for Figure 16b immediately follows.

Data for Figure 16b

Date	Total change	Net flows	Valuation changes	Estimated gap
1995	178	51	70	57
1996	219	64	97	58
1997	163	43	70	50
1998	286	106	140	41
1999	545	69	430	47
2000	-148	94	-269	28
2001	-279	97	-373	-2
2002	-259	5	-342	78
2003	726	107	492	127
2004	481	74	330	77
2005	757	132	388	238
2006	1011	128	750	134

contributions are almost uniformly positive and at times have been quite large. For example, in 2002 and 2003, the gaps explain most of the total change in U.S. holdings of foreign bonds (Figure 16a). Whereas the reported flow data for 2002 and 2003 indicated that U.S. investors on net sold foreign bonds over this two-year period while positive valuation effects provided some offset, holdings of foreign bonds as reported in the December 2003 survey were considerably larger than could plausibly be explained by these factors. Subsequent investigation indicated that this miss was primarily the result of underreporting of foreign bonds newly issued in the United States, suggesting that in this case the gap at least in part should be attributed to missed flows.²⁴ In general, the fairly persistent positive gaps for estimates of U.S. claims may reflect a greater difficulty in measuring U.S. cross-border acquisitions of foreign securities, as well as the inherent difficulty in making accurate valuation adjustments for the sizable, diverse portfolio of foreign securities held by U.S. investors.

7 Conclusions and further research

Using an approach based on Thomas, Warnock and Wongswan, we provide monthly estimates of U.S. cross-border securities positions combining periodic comprehensive surveys of holdings with the monthly TIC cross-border securities transactions data. We extend the dataset used in earlier work, make somewhat more comprehensive adjustments to the transactions data and enhancements to prices used to make valuation adjustments, and clarify the calculations involved. Although we are unable to identify a satisfactory statistical method for extrapolating estimates beyond the most recent survey values, we arrive at a seemingly workable ad hoc approach for updating estimates through the current transactions data.

Our explorations suggest that further improvements could be made to our valuation adjustments, especially for those applied to foreign holdings of U.S. equities and U.S. corporate bonds. Such improvements not only would improve our ability to estimate holdings for dates beyond the last survey, but would also enhance our understanding of the returns foreigners earn on their portfolio investments in the United States relative to the returns U.S. investors earn abroad, an issue of considerable current interest.

In future work we also hope to further explore further the statistical basis for estimation, possibly using bootstrapping or nonparametric methods and accounting for other factors such as trading volume to better explain observed gap errors. Additionally, we hope to extend our statistical analysis to the U.S. claims data to better understand the sources of the persistent positive errors in estimating U.S. holdings of foreign securities.

References

C.C. Bertaut, W.L. Griever, and R.W. Tryon (2006), "Understanding U.S. Cross-Border Securities Data", Federal Reserve Bulletin, vol. 92 (Spring), pp. 59-75.

W. L. Griever, G. A. Lee, and F. E. Warnock (2001), "The U.S. System for Measuring Cross-Border Investment in Securities: A Primer with a Discussion of Recent Developments", Federal Reserve Bulletin, vol. 87 (October), pp. 633-50.

C. P. Thomas, F. E. Warnock, and Jon Wongswan, "The Performance of International Portfolios", International Finance Discussion Papers Series 2004-817 (Washington: Board of Governors of the Federal Reserve System, October), www.federalreserve.gov/pubs/ifdp/2004/817/default.htm.

F.E. Warnock and C. Cleaver, "Financial Centers and the Geography of Capital Flows", International Finance Discussion Papers Series 2002-722 (Washington: Board of Governors of the Federal Reserve System, April), http://www.federalreserve.gov/pubs/ifdp/2002/722/ifdp722.pdf.

F.E. Warnock and M. Mason, "The Geography of Capital Flows: What We Can Learn from Benchmark Surveys of Foreign Equity Holdings", International Finance Discussion Papers Series 2000-688 (Washington: Board of Governors of the Federal Reserve System, December), http://www.federalreserve.gov/pubs/ifdp/2000/688/ifdp688r.pdf

Appendix A Adjusting for sales of asset-backed securities, stock swaps, and transactions costs

Asset-backed Securities

As discussed in Bertaut, Griever, and Tryon (2006), an important adjustment to estimates of foreign holdings of U.S. debt securities is to account for repayments of principal on asset-backed debt securities (ABS).²⁵ Although foreign purchases and sales of agency and corporate ABS are included in the monthly TIC transactions data, the periodic repayments of principal prior to redemption are not reported, because these repayment streams do not pass through the brokers and dealers primarily responsible for reporting securities transactions. Thus, using the as-reported data on agency and corporate debt transactions without accounting for these repayments will tend to overstate foreign net acquisitions of asset-backed debt securities, because the monthly data will include any new purchases of ABS made to offset principal paydowns, but not the paydowns themselves. The estimated adjustments for ABS repayments can be sizable: for 2005, these estimates reduce net purchases of U.S. government agency bonds by nearly $63 billion and net purchases of corporate bonds by nearly $81 billion.

To adjust the transactions data for ABS repayments flows, we use the estimates of monthly ABS paydowns calculated by the Federal Reserve Bank of New York and available on the TIC website. Repayment rates are derived from the underlying security-by-security data from the U.S. liabilities surveys, reported monthly repayment rates by Fannie Mae and Freddie Mac (for agency ABS), and individual security factor values from Bloomberg Online Data services (for corporate ABS). These estimates are available from July 2002 forward on the TIC web site at http://www.treas.gov/tic/; for agency bonds, estimates of ABS repayment flows for securities held by foreign official institutions are also provided. Details of the methodology are also available on this site.²⁶ Although asset-backed principal repayments are potentially a concern on the asset side as well as the liabilities side, it does not appear that such repayment flows are a serious omission at present. Although holdings of ABS securities have grown from less than 4 percent as of December 2001 to more than 12 percent as of December 2005, many of these foreign ABS securities replace repaid principal, and thus monthly net purchases are not likely to be overstated to the same degree.

Stock swaps

Another important omission in the TIC data, noted by Thomas, Warnock and Wongswan, is that the TIC S data fail to capture U.S. acquisitions of foreign stock and foreign acquisitions of U.S. stock that arise from stock swaps associated with corporate mergers or takeovers. When a foreign company acquires a U.S. company and the deal is financed in part through a stock swap, U.S. residents who held stock in the target company become holders of foreign equity. Likewise, if a U.S. company acquires a foreign company, a stock swap can increase foreign holdings of U.S. equity. These stock swaps are omitted from the TIC data, but they are reported in the BEA's quarterly balance of payments statistics. Although merger activity has tapered off in recent years, stock swaps previously were an important source of financing for U.S. acquisitions of foreign equity. For example, in 2000, U.S. residents acquired $13 billion in foreign equity through net purchases but $80 billion through stock swaps associated with foreign acquisitions of U.S. companies.

Following Thomas, Warnock and Wongswan, we adjust net transactions of both U.S. and foreign equity to account for acquisitions of equity resulting from stock swaps. Data on the financing of corporate mergers and takeovers from Security Data Corporation are used to distribute BEA's quarterly statistics of stock swaps by month, and these monthly estimates are available on the TIC website.²⁷ The published estimates are reported for all foreign acquisitions of U.S. equity and for all U.S. acquisitions of U.S. equity that are the result of stock swaps. We use the same Security Data Corporation financing information which details the country of incorporation of both target and acquiring companies to distribute the stock swap estimates by country.

Transactions Costs

Because the TIC S data include transactions costs in any cross-border financial transaction, using the as-reported transactions data can introduce an additional source of error in constructing position estimates. For example, if a foreign resident purchases $10,000 in U.S. securities with a brokerage fee of 1 basis point, the total dollar amount remitted by the foreign purchaser - and thus reported on the TIC S - will be $10,001, although foreign holdings of U.S. securities will increase by only $10,000 (assuming no valuation changes). However, this source of error is likely to be fairly small compared to the omission of stock swaps and ABS repayment flows. Although the volume of U.S. cross-border transactions has increased dramatically, the fees and commissions on transactions in U.S. securities have also declined significantly in recent years, and thus the current fraction of net purchases accounted for by transactions costs is likely to be quite small. Our methodology allows us to ignore transactions costs in constructing position estimates by incorporating them with the transactions measurement error.

Appendix B Adjusting for Valuation Changes

We use information from the detailed surveys of holdings to determine appropriate price indexes to apply to account for valuation changes. For U.S. Treasury securities, the choice of an appropriate index is fairly straightforward, because foreign holdings of U.S. Treasuries are reasonably well represented by a standard index of Treasury securities. Similarly, we us a standard U.S. equity price index to account for valuation changes to foreign holdings of U.S. equities, although this pricing choice may overstate the extent of valuation changes because the definition of "equity" also includes mutual funds.²⁸ For U.S. agency bonds and corporate bonds, we account for the growing role of asset-backed securities in foreign portfolios by creating weighted average indexes of ABS and conventional agency debt and of ABS and conventional corporate debt, where the weights are derived from the proportions of ABS and conventional debt as reported in the surveys.²⁹ For U.S. corporate debt, we further account for the proportion of debt that is issued in foreign currency: As of June 2005, roughly 25 percent of U.S. corporate debt was denominated in foreign currency, largely in euro. Our conventional debt index itself is a weighted average of foreign currency and dollar-denominated U.S. corporate debt, where the foreign currency component is constructed from indexes of euro-, sterling-, and yen-issue corporate bonds.

Accounting for valuation effects for U.S. holdings of foreign securities is considerably more complex, because the U.S. portfolio of foreign stocks and bonds includes securities issued by many different countries, which greatly compounds the estimation problem. To estimate valuation changes for foreign securities, we usually use individual country equity or bond price indexes, taking into account the currency composition of U.S. holdings.³⁰

Pricing holdings of securities issued by offshore financial centers - such as Bermuda and the Cayman Islands - provides a special challenge, because there are no appropriate country equity or bond price indexes to apply. U.S. holdings of securities issued in offshore centers, especially those in the Caribbean, in large part consist of equity or debt of companies established or reincorporated in these locations.³¹ Securities issued in offshore financial centers may also be issued by entities controlled by onshore corporations, including "special-purpose vehicles" (SPVs).³² U.S. holdings of financial center equity tend to be highly concentrated in the shares of a few corporations, most of which are issued in dollars and trade primarily on U.S. exchanges, rather than the exchange of the country of incorporation. Where U.S. holdings of financial center equity are heavily concentrated in shares of a few specific corporations, we construct customized indexes of equity based on the prices of the shares held by U.S. investors.

Holdings of long-term debt issued in offshore financial centers present a different challenge. In this case, U.S. holdings consist largely of debt securities issued through SPVs, especially those established in the Cayman Islands. The growth of securities issued through such entities accounts for much of the increase in Cayman Island debt held by U.S. investors: Such debt amounted to about 2 percent of U.S. investors' holdings of foreign bonds in 1997 but grew to nearly 12 percent by 2005. The majority of these bonds are U.S. dollar-denominated, and an increasing share is asset-backed: asset-backed securities accounted for nearly 60 percent of U.S. residents' holdings of Cayman Island bonds in 2005. We estimate valuation changes in Caribbean financial center long-term debt by applying weighted averages of world sovereign, U.S. corporate, and U.S. corporate asset-backed securities indexes.

Appendix C Allocation of the residual measurement error across months between surveys

Beginning with an initial survey position, an estimate of the position at a future date can be calculated as

(A1) $\hat{S}_{t} =S_{0} (1+\hat{\pi}_{0,t} )+\sum\limits_{i=1}^{t} {\hat {N}_{i} } (1+\hat{\pi}_{i,t} )$

where $S_{0}$ is the latest survey value for a given country, security, and holder; $\hat{S}_{t}$ is the estimated position at time >, $\{\hat {N}_{i} \}$ is the sequence of net flows from time 1 to t and $\hat{\pi }_{i,t}$ is the rate of increase in the price of security S over the period i to t, with $\pi_{i,i} =0$ .

In this case, only $S_{0}$ is actually known; $\hat{N}_{i}$ and $\hat{\pi }_{i,t}$ are observed with error and $\hat{S}_{t}$ is an estimate. When , the date of the next survey, $S_{T}$ is known, and we can define the "gap" between the actual and the estimated survey positions:

(A2) $G_{T} =S_{T} -\hat{S}_{T}$

We define the measurement errors in security prices $(\varepsilon_{t} )$ as

(A3) $(1+\pi_{t} )=(1+\hat{\pi}_{t} )(1+\varepsilon_{t} )$

where $\varepsilon_{t}$ is the multiplicative error in observing the true monthly valuation change, $\pi_{t}$ . We correct net transactions for transactions costs and other errors:

(A4) $N_{t} =(1+\beta_{t} )\hat{N}_{t}$

where $\beta_{t}$ includes both effects in multiplicative form. (We assume that both transactions costs and other measurement errors are equal percentages of both purchases and sales, so that $\beta_{t}$ may be applied directly to net transactions.) The laws of motion for $S_{t}$ and $\hat{S}_{t}$ are

(A5) $S_{t} =S_{t-1} (1+\pi_{t} )+N_{t}$

$\displaystyle \hat{S}_{t} =\hat{S}_{t-1} (1+\hat{\pi}_{t} )+\hat{N}_{t}$

Using (A3) - (A5) we define the gap between the (unobserved) actual position and the observed position at time as

(A6) $G_{t} \equiv S_{t} -\hat{S}_{t}$

         $\displaystyle =S_{t-1} (1+\pi_{t} )+N_{t} -\hat{S}_{t-1} (1+\hat{\pi}_{t} )-\hat{N}_{t}$
         $\displaystyle =S_{t-1} (1+\hat{\pi}_{t} )(1+\varepsilon_{t} )+(1+\beta_{t} )\hat{N}_{t} -\hat{S}_{t-1} (1+\hat{\pi}_{t} )-\hat{N}_{t}$
         $\displaystyle =\hat{S}_{t-1} (1+\hat{\pi}_{t} )\varepsilon_{t} +\beta_{t} \hat{N}_{t} +G_{t-1} (1+\hat{\pi}_{t} )(1+\varepsilon_{t} )$

Recursively substituting for , we have

(A7) $G_{t} =\sum\limits_{i=1}^{t} {\left[ {\hat{S}_{i-1} (1+\hat{\pi}_{i} )\varepsilon_{i} +\beta_{i} \hat{N}_{i} } \right] \prod\limits_{j=i+1}^{t} {(1+\hat{\pi}_{j} )(1+\varepsilon_{j} )} }$

Note that for the summation is from 1 to :

(A8) $G_{T} =\sum\limits_{t=1}^{T} {\left[ {\hat{S}_{t-1} (1+\hat{\pi}_{t} )\varepsilon_{t} +\beta_{t} \hat{N}_{t} } \right] \prod\limits_{i=t+1}^{T} {(1+\hat{\pi}_{i} )(1+\varepsilon_{i} )} }$

and the period contribution to the total end of period gap can be defined as the -th element in the summation in (A8), or:³³

(A9) $g_{t} \equiv\left[ {\hat{S}_{t-1} (1+\hat{\pi}_{t} )\varepsilon_{t} +\beta_{t} \hat{N}_{t} } \right] \prod\limits_{i=t+1}^{T} {(1+\hat{\pi}_{i} )(1+\varepsilon_{i} )}$

The ratio of (A9) to (A8) is the share of period in the total gap at time :

(A10) $\lambda_{t} =\frac{g_{t} }{G_{T} }=\frac{\left[ {\hat{S}_{t-1} (1+\hat{\pi}_{t} )\varepsilon_{t} +\beta_{t} \hat{N}_{t} } \right] \prod\limits_{i=t+1}^{T} {(1+\hat{\pi}_{i} )(1+\varepsilon_{i} )} } {\sum\limits_{t=1}^{T} {\left[ {\hat{S}_{t-1} (1+\hat{\pi}_{t} )\varepsilon _{t} +\beta_{t} \hat{N}_{t} } \right] \prod\limits_{i=t+1}^{T} {(1+\hat{\pi }_{i} )(1+\varepsilon_{i} )} } }$

The quantity $\lambda_{t} G_{T}$ is the amount of the end-of-period gap attributable to month t.³⁴ adjusting for the valuation change between t and T gives us the quantity to be added to the measured position at time t. Our final estimated position $\tilde{S}_{t}$ is the observed position ( $\hat{S}_{t} )$ plus the allocated share of the gap:

(A11) $\tilde{S}_{t} =\tilde{S}_{t-1} (1+\hat{\pi}_{t} )+\hat{N}_{t} +\frac{\lambda_{t} G_{T} }{\prod\limits_{i=t}^{T} {(1+\hat{\pi}_{i} )(1+\varepsilon_{i} )} }$

By evaluating (A11) at t = T, it is straightforward to show that $\tilde{S}_{T} =S_{T}$ , and since (A11) is the law of motion for $\tilde{S}_{T}$ , the sequence $\tilde{S}_{T}$ is "survey consistent" in the sense of Thomas, Warnock, and Wongswan (2004).

Appendix D Estimated positions in downloadable format

We provide time series of monthly estimates of foreign holdings of U.S. securities and U.S. holdings of foreign securities, along with the basic decomposition into monthly flows, valuation changes, and estimated gap contributions as described in section 6. Estimates are available by security type and country for the dates December 1984-June 2006 (foreign holdings of US securities) and December 1994-December 2006 (US holdings of foreign securities). We report estimated positions for between-survey values only, and leave the decision of how to forecast future values to the user's discretion.

The time series are stored in three comma-delimited text files:

ticdata.liabilities.ftot.txt

ticdata.liabilities.foi.txt

ticdata.claims.txt

where 'ftot' and 'foi' denote total foreign holdings and holdings by foreign official institutions. The three files are combined into a zip file

ticdata.zip

In each file the time series are arranged in columns; the rows are ordered by country code and by date. The first row contains the column names, as follows:

ticdata.liabilities.ftot.txt

date (mm/dd/yyyy)
country code (ccccc)
country name

ftot_*_survey_pos	reported survey values
ftot_*_adj_net_flow	adjusted monthly net flows
ftot_*_adj_val_chg	estimated monthly valuation changes
ftot_*_monthly_gap	estimated monthly gap contributions
ftot_*_est_pos	estimated positions

where * =

agcy	U.S. agency bonds
corp	U.S. corporate bonds
stk	U.S. stocks
treas	U.S. treasury bonds

ticdata.liabilities.foi.txt is similar, with the obvious substitution of foi for ftot. (Note that only data for the total of all foreign official holders is available). The claims file is for total claims for two security types, stocks (stk) and bonds (bond):

ticdata.claims.txt

date (mm/dd/yyyy)
country code (ccccc)
country name

us_*_survey_pos	reported survey values
us_*_adj_net_flow	adjusted monthly net flows
us_*_adj_val_chg	estimated monthly valuation changes
us_*_monthly_gap	estimated monthly gap contributions
us_*_est_pos	estimated positions

Footnotes

* Division of International Finance, Board of Governors of the Federal Reserve System. All views expressed are the authors' own and not necessarily those of any member of the Board of Governors or its staff. We thank Charles Thomas, Stephanie Curucuru, Jaime Marquez, Eric Hjalmarsson, Daniel Beltran, David Bowman, and other participants at an International Finance Division workshop for helpful comments and suggestions; all mistakes are our own. We thank Zachary Kurtz and Gary Maguire for research and programming assistance. Return to text

1. Surveys of foreign holdings of U.S. securities were also conducted in 1974 and 1978, but the published data from these surveys is limited. The publicly available data for 1978 include total foreign holdings by country of U.S. equities and long-term debt securities, but with no breakdown by country by type of debt security. The survey for 1974 includes a breakdown between U.S. government debt securities and corporate debt securities, but data by country are available only for private foreign investors. Return to text

2. Survey data are collected and reported for approximately 200 countries, but our sample is limited by the countries for which a time series of transactions data are also available. Return to text

3. We focus on estimates of holdings of long-term securities in this paper because timely estimates of cross-border holdings of short-term securities are reported elsewhere by the TIC system. For further information on the TIC cross-border securities data, refer to Bertaut, Griever, and Tryon (2006). Return to text

4. Because transactions data for the United Kingdom are not available separately from data for the Channel Islands and the Isle of Man prior to 2001, the concept "United Kingdom" throughout this paper refers to the consistently defined broader aggregate. Return to text

5. Although the liabilities surveys are thought to be comprehensive in their coverage of total foreign holdings of U.S. securities, an important caveat to the accuracy of the surveys is that the country attribution of foreign holdings can be distorted by "custodial bias". If a foreign investor acquires U.S. securities and holds them with a foreign custodian bank, the foreign bank will typically employ a U.S. custodian bank to facilitate settlement and custody operations. When the U.S. bank reports these holdings on the U.S. liabilities survey, it typically will only know that the securities are held at the foreign custodian bank, and will not know the nationality of the actual foreign owner. Thus, the U.S. liabilities surveys tend to attribute very large foreign holdings to countries that are major custodial, securities management, or depository centers, such as Belgium, the Cayman Islands, Luxembourg, and Switzerland. This "custodial bias" is separate from the "financial center" bias in the monthly transactions data. For more information on the problem of "custodial bias" and a discussion of how the IMF's CPIS asset surveys can help us gain a better understanding of the ultimate foreign owners of U.S. securities, see Bertaut, Griever, and Tryon (2006). Return to text

6. Strictly speaking, we assume that all observations are made on the last day of the month, and thus we do not adjust for valuation changes on flows over the course of the month. Return to text

7. Negative estimates reflect a basic inconsistency between the annual surveys and the monthly transactions data that cannot easily be reconciled. In some instances, the true position may in fact be negative if foreign investors hold short positions in the security in question. Short sales of securities can, and presumably are, included in the net transactions data, but short positions are not collected in the annual surveys. Since short sales cannot be identified in the data, we impose nonnegativity as described. Return to text

8. Equation (5) also implicitly includes $\varepsilon_{t}$ . Return to text

9. This approach differs from that of Thomas, Warnock and Wongswan, who do not model the measurement error explicitly and instead assume that it is proportional to gross transactions. Return to text

10. Our overall approach closely follows Thomas, Warnock and Wongswan, but note that equation (11) can be evaluated directly (period by period), without the need for a nonlinear method as in TWW. Return to text

11. Estimates for all foreigners are constructed by applying this methodology to survey values of holdings by all foreigners and using adjusted net transactions by all foreigners. For between-survey dates, these estimates will differ somewhat from the sum of estimated holdings of each individual country for two reasons. First, imposing non-negativity is more likely to bind at the country level than for all foreigners. Thus, constructing total holdings by summing individual country values may result in a somewhat larger estimate because each individual country's estimate is bounded below at zero. The second reason is the significant fraction of securities measured in the surveys for which the country of owner cannot be identified. Most of these securities are unregistered or "bearer" securities. Although bearer securities generally cannot be issued in the United States, U.S. firms can and do issue such securities aboard. Because the owners of these securities need not make themselves known, little or no information is typically available on who these foreign owners are. Such securities are reported in the surveys under "country unknown." Holdings by "country unknown" have been sizable at times, amounting to more than $460 billion or about 40 percent of all foreign holdings of U.S. corporate bonds in June 2002. Because we have no information on what share of transactions in U.S. corporate bonds reflect transactions in bearer securities, we estimate between-survey holdings of "country unknown" by linear interpolation. Return to text

12. As long as we assume that the "naïve" positions (reflecting both the magnitude of the previous survey position and the evolution of net flows and valuation changes) tell us something about the actual "end-of-period" positions, some assumptions have to be made about the about the end-of period gaps. An alternative approach might be to simply forecast the end-of-period position itself, based (for example) on trend growth in the position, and thus ignoring the contributions from net flows and valuation changes. This approach may be appropriate for especially noisy countries such as those with large financial center bias, but seems unappealing for estimating positions more generally. Return to text

13. We scale by the estimated position rather than the actual survey value so that when constructing estimates for current holdings, when we have only the estimated position to work from, the results will be comparable. Because we restrict the estimated positions to be non-negative, this choice also has the result that the minimum the scaled gap can be is -1, the result if the actual survey value is small relative to the estimated position (at the limit, the survey position is zero, and the estimated position is positive). In contrast, there is no upper bound to the maximum scaled gap. Return to text

14. Formal tests for normality confirm this intuition. Another possibility is that the gaps should be scaled by a quantity other than the estimated position that we used. Unfortunately we were unable to identify a scale factor that improved the distribution of the gaps in any significant way. Return to text

15. Discouraging as these results may be, fit is notably better than for the full sample including outliers: conducting the same regression exercise over the full sample generates an adjusted R² of -.005 and a RMSE of 40.63. Return to text

16. When we have data for the 2007 survey, this procedure would imply that we use the average of five survey gaps. We may instead use a moving average of the four most recent gaps, to allow for possible shifts over time. Return to text

17. Foreign official holdings of asset-backed agency securities have grown from $4.3 billion in June 2002 to $117.7 billion in June 2006, and now account for roughly 38 percent of all agency securities held by official institutions, and 30 percent agency ABS held by all foreign investors. Corporate ABS holdings by foreign official investors are much smaller: $29.7 billion in June 2006, or about 5 percent of all foreign holdings of corporate ABS. Estimates of monthly repayment flows on foreign official holdings of agency ABS are available on the TIC website. Return to text

18. Consider the example of a foreign official institution that acquires a U.S. security, such as a Treasury bond, from a private foreign entity on a foreign securities exchange and then has the security moved to be held in custody in the United States at the Federal Reserve Bank of New York (FRBNY). In this case, reported holdings of Treasury securities by foreign official institutions will increase, but no corresponding TIC-reported foreign official purchase will be recorded because the acquisition by the foreign official institution from another foreigner is not a U.S. cross-border transaction: It is a foreign-to-foreign transaction. Because the FRBNY is a reporter on the annual surveys, the increased holdings will then be measured in the next survey. When the private foreigner first acquired the Treasury security, a U.S. cross-border transaction would have been reportable in the TIC system. However, it would not have been recorded as a foreign official purchase, nor would it necessarily have been recorded in the same calendar month or against the same country as was the movement into U.S. custody. For more information, including a comparison of changes in Treasury and agency securities held at the Federal Reserve Bank of New York for official accounts, refer to TIC FAQ number 10, www.treas.gov/tic/faq1.html. Return to text

19. We do not have a comparable "country unknown" problem for our claims surveys, because the security-level detail allows us to correctly identify the country of issuer even if the security is unregistered. Likewise, the collection of security-level data in our asset surveys means that we do not have the same "custodial bias" that is present in our liabilities surveys. Return to text

20. In its presentation of the IIP, the Bureau of Economic Analysis decomposes the change in the year-end positions in a similar fashion, separating the annual contributions into financial flows, valuation effects that in turn are separated into price changes and exchange rate changes, and "other factors." Return to text

21. We also bear in mind that our definition of "flows" includes transactions costs. Return to text

22. Because the more recent liabilities surveys are conducted as end-June, the effect of a negative coefficient for the 2002 survey would show up as contributing to the negative gap contributions for both 2001 and 2002, and likewise that for the 2003 survey will contribute to the gaps for both 2002 and 2003. Return to text

23. And in fact although the BEA does allow much of the "gap" arising from the results of a new survey to show up in the estimated "other" factors in their presentation of changes underlying the IIP, they also make some judgmental allocations of the gap to adjust financial flows and valuation changes. Return to text

24. In the international financial transactions accounts, the BEA reports U.S. net purchases of foreign bonds of roughly $30 billion in each of 2002 and 2003, indicating that they did make a judgmental allocation of at least part of the gap to the net flow contribution. Return to text

25. Asset-backed securities are securities backed by pools of assets, such as residential home mortgages, which give the security owners claims against the cash flows generated by the underlying assets. Unlike most other debt securities, ABS typically repay both principal and interest on a regular basis, reducing the principal outstanding with each payment cycle. Foreign-held asset-backed agency securities are largely mortgage-backed securities issued by the Federal National Mortgage Association (Fannie Mae) and the Federal Home Loan Mortgage Association (Freddie Mac). Corporate asset-backed securities include mortgage-backed securities issued by non-government corporations as well securities backed by pools of assets such as auto loans and credit card receivables. Return to text

26. It should be noted that the published ABS repayment flow estimates are subject to revision as information from a new liabilities survey becomes available. For example, if the new survey indicates that foreign ABS holdings are a larger share of agency or corporate securities than in the previous survey, foreign holdings of ABS in both the inter-survey period and in months following the survey are likewise presumed to be larger, generating larger ABS repayment flows. Return to text

27. These monthly estimates from January 2000 forward are available at www.treas.gov/tic/swapstk.html; in adjusting for stock swaps in our estimates, we include additional estimates of swaps back to the mid-1980s. Return to text

28. Roughly 10 percent of U.S. liabilities classified as equities are foreign holdings of U.S. mutual funds. While many of the funds themselves are equity funds, others are hybrid funds, bond funds, or money market funds. Return to text

29. Detail on ABS and non-ABS securities are first reported in the June 2002 survey. At that time, ABS accounted for 25 percent of long-term agency debt held by foreigners and 15 percent of long-term corporate debt held by foreigners. These weights increased to 28 and 19 percent, respectively, in 2004, to 33 and 26 percent in 2005, and to 39 and 29 percent in 2006. We linearly interpolate the weights between survey values. For estimates of holdings before June 2002, we use standard indexes of conventional agency and corporate debt. Return to text

30. A surprisingly large portion (more than 75 percent in December 2005) of U.S. holdings of foreign bonds are dollar-denominated, including 89 percent of bonds issued in Luxembourg, 84 percent of Australian bonds, 83 percent of U.K. bonds, and between 71 percent and 77 percent of Canadian, Dutch, and Irish bonds. We use weighted averages of local currency bond indexes and Eurodollar bond indexes to estimate valuation changes for holdings of bonds issued by most industrial countries. For debt issued by emerging market countries, we use country-specific weighted averages of EMBI+ indexes and local-currency bond indexes. Although the majority of emerging market debt held by U.S. investors is dollar-denominated, the share of local-currency debt has picked up in recent years. Return to text

31. For balance of payments purposes and thus within the TIC system, an institution is considered to be resident in the country in which it is incorporated or otherwise legally created. Thus, when these entities issue securities, they will be attributed to the country of the offshore financial center rather than to the country of the onshore parent corporation, even though the onshore parent corporation may be understood to be the ultimate obligor. In addition, some companies have reincorporated from their country of origin to offshore financial centers for tax purposes. Although the reincorporation probably has little or no effect on their locus of activity, securities issued by these companies will now be attributed to the country of reincorporation. Equity issued in financial centers accounts for a growing percentage of the U.S. portfolio of foreign assets: In 1997, U.S. holdings of equity issued by Caribbean offshore financial centers amounted to $48 billion, or roughly 4 percent of foreign equity held by U.S. investors. By the end of 2004, these amounts had grown to $277 billion, or nearly 11 percent of foreign equity held. Return to text

32. A special-purpose vehicle is a legal entity that may be created in an offshore financial center (OFC) to engage in financial activities in a more favorable tax environment. An onshore corporation establishes an SPV in an offshore center to engage in a specific activity, such as the issuance of asset-backed securities. The onshore corporation may assign a set of assets to the offshore SPV (for example, a portfolio of mortgages, loans, or credit card receivables). The SPV then offers a variety of securities to investors based on the underlying assets. The SPV, and hence the onshore parent, benefit from the favorable tax treatment in the OFC. Return to text

33. The decomposition of $G_{T}$ in (A9) and (A10) is not unique, because of the interaction between $\varepsilon_{i}$ and $\varepsilon_{j}$ , $i\ne j$ in the product on the right-hand side. These interactive effects could plausibly be assigned either to $g_{i}$ or to $g_{j}$ . However, this definition is intuitively appealing - it represents the direct effect of the time measurement errors on the time gap when they have been revalued at time . Return to text

34. Obviously, the definition of $\lambda_{t}$ in (10) requires $G_{T} \ne0$ . If this condition is not met, there is no gap, no measurement error, and we do not need to define the weights $\lambda_{t}$ at all. Return to text

This version is optimized for use by screen readers. Descriptions for all mathematical expressions are provided in LaTex format. A printable pdf version is available. Return to text

Monthly Estimates of U.S. Cross-Border Securities Positions*

1 Introduction

2 The discrepancy between survey positions and transactions-based positions estimates

3 Estimating monthly positions between surveys

4 Estimating positions after the last survey

4.1 Summary statistics

4.2 Forecasting the survey gaps

4.3 Estimating positions for foreign official investors

5 Estimating positions on the U.S. claims side

6 Decomposing changes between surveys into flows, valuation changes, and other factors

7 Conclusions and further research

References

Appendix A Adjusting for sales of asset-backed securities, stock swaps, and transactions costs

Appendix B Adjusting for Valuation Changes

Appendix C Allocation of the residual measurement error across months between surveys

Appendix D Estimated positions in downloadable format

Footnotes

Monthly Estimates of U.S. Cross-Border Securities Positions^*