Flow of funds household balance sheet background


9 Jun 2011.

The Flow of Funds report from the Federal Reserve provides, in tables B.100 and L.100, the only serious, public, regularly revised estimate of the household sector balance sheet. Derivation of the numbers is complex and involves some guesswork, so all results should be taken with a grain of salt. However the trends are most likely correct.


Error bars are impossible to estimate (9 Jun 2011)

  • The numbers are derived from a very large number of sources, some of uncharacterized precision
  • Some parts of the derivation involve rule-of-thumb guesses (see on mortgage debt below), and
  • Some important calculations are not documented at all.
  • One research paper from the Federal Reserve states, “In the FFA [Flow of Funds Accounts], financial assets and liabilities of the household sector are largely derived as residuals [i.e. totals are calculated for the business and government sectors, and then any leftover is assigned to the household sector] … the complex structure of the FFA, with all sectors, in a sense, leading to the household sector, and the vast disparate sources that are used as inputs–about 3500 data series are currently used to compile the Flow of Funds Accounts–make calculating even the most simplistic standard error a daunting, if not impossible, task.”

Housing value (9 Jun 2011) The Fed benchmarks aggregate housing value to the American Housing Survey, which is conducted by Census. Census surveys a total of about 55k housing units (owned and rented) every two years, and supplements this with smaller, staggered surveys of 47 metro areas. Owners are asked what they paid for the house. The Fed extrapolates this to the entire market and then, quarterly, (1) adds new residential construction totals and (2) adjusts the value of construction in place via the Loan Performance House Price Index. It is hard to see how one could get a better estimate, but it is also quite unclear what the precision might be.

US mortgage debt outstanding background (9 Jun 2011) Table L.218 of the Flow of Funds provides a useful overview of total home mortgage funding and how it is split between different classes of lenders. This includes all debt secured by 1-4 family homes – purchase loans, lines of credit, home equity loans, and construction and development loans. The breakdown is by the holder of the original loan, so when an MBS is issued based on a pool of loans, the pool, and not the owner of the MBS, is considered the holder.

For total home mortgage debt of the household sector, one can use either line 1, which is reliable, but too large, because it includes construction and development loans, or line 2, which should be the right total, but whose calculation includes some rough guesses.

Three high level categories suffice to give a reasonably clear picture of how mortgage funding is being supplied: (1) loans held directly on private balance sheets (at banks, thrifts, credit unions, finance companies and REITs, (2) loans backed by the government (grouping (a) the pools underlying securities issued by Fannie Mae, Freddie Mac and Ginnie Mae with (b) unbundled loans held directly on the balance sheets of Fannie and Freddie), and (3) private-label ABS. These together cover over 95% of the total market.

Clippings below were used in the construction of this page

Covered through 21 Dec 2009.

The SCF provides an independent check of the FFA

Oct 2000. Federal Reserve research paper.

“A Comparison of the Household Sector from the Flow of Funds Accounts and the Survey of Consumer Finances. Rochelle L. Antoniewicz”

“The most widely used source of aggregate data for U.S. household balance sheets is the time series data from the Flow of Funds Accounts (FFA). In the FFA, financial assets and liabilities of the household sector are largely derived as residuals because reports on the balance sheet activities of households are generally not available, except intermittently. In other words, the FFA starts with known economy wide totals for individual transaction categories and then amounts reported to be held by other sectors are deducted, leaving the household sector with the remainder. For most transaction categories, such as home mortgage debt and time deposits, this method seems reasonable because the household sector is the largest holder. Yet uncertainty about the accuracy of the asset and liability estimates in the FFA household sector remains and at times, the FFA estimates have been in question because of their residual nature.

This paper addresses some of these questions by comparing figures on selected assets and liabilities from the FFA household sector with survey-based estimates from the 1989, 1992, 1995, and 1998 Surveys of Consumer Finances (SCF)–the most comprehensive survey on household wealth. …

the complex structure of the FFA, with all sectors, in a sense, leading to the household sector, and the vast disparate sources that are used as inputs–about 3500 data series are currently used to compile the Flow of Funds Accounts–make calculating even the most simplistic standard error a daunting, if not impossible, task. …

[earlier] studies generally did not fully adjust the FFA and SCF measures to place them on a definitionally equivalent basis. …

The FFA and SCF estimates of total liabilities differ by 1.4 percent in 1989, 0.4 percent in 1992, 10.4 percent in 1995, and 4 percent in 1998. Much of the wider discrepancy between the FFA and SCF liability estimates in 1995 and 1998 owes to a significant difference between the two measures of consumer credit. After nearly exact matches in 1989 and 1992, the two measures diverged, particularly in 1995, with the FFA estimate of consumer credit growing much faster than the SCF estimate. On the asset side, the differences between the two measures of total assets are 2.1 percent in 1989, 7.2 percent in 1992, 9.3 percent in 1995, and 0.9 percent in 1998. For some asset categories, such as, mutual fund shares, owner-occupied real estate, and checkable deposits, the FFA and SCF estimates are very close in 1989 and 1992, but move apart in 1995 and 1998. For other assets, such as saving deposits and corporate equity, considerable differences, although smaller than those documented in previous studies, remain due to unresolved definitional issues or measurement error in either data set. …

Before one can compare the SCF and FFA asset estimates, several adjustments to both measures are necessary. The most crucial adjustments account for the broader inclusion of assets in the FFA and the different treatments of IRA/Keogh accounts and employer-sponsored private pension assets in the FFA and the SCF. The household asset estimates as reported in the FFA include assets of nonprofit organizations, unit investment trusts, and investment management accounts, none of which are reported in directly held assets in the SCF.5 These institutional assets account for between 5-1/2 percent and 7-1/2 percent of the FFA household sector's total financial assets over the 1989 to 1998 period. …

if adjusted for definitional differences, the SCF and FFA estimates of home mortgage debt are fairly close, although they do move apart a bit in 1995 and 1998, but still remain to within close to one SCF standard error. …

After careful adjustments for conceptual and definitional differences in the FFA and SCF transaction categories, I find that the FFA and SCF estimates for total liabilities and total assets are extremely close in 1989. Indeed, the 1989 FFA estimates of home mortgage debt, consumer credit, U.S. government securities, corporate and foreign bonds, municipal securities, mutual fund shares, publicly traded corporate equity, money market mutual funds, equity in noncorporate business, and owner-occupied real estate are all within one standard error of the SCF estimates. The match up between the FFA and SCF estimates becomes progressively worse for the remaining survey years. By 1998, only the FFA estimates of home mortgage debt, municipal securities, and equity in noncorporate business are within one standard error of the SCF estimates.

The main trouble between the SCF and FFA estimates primarily lies in the consistent and puzzling offsetting differences between the SCF and FFA estimates for time and saving deposits and the value of closely held corporate equity. The FFA shows higher time and saving deposits than the SCF, while the SCF shows higher closely held corporate equity.”

Flow of Funds background

2000. Guide to the Flow of Funds Accounts.

Published by the Board of Governors of the Federal Reserve System.

Household balance sheet numbers usually reported in the news are from the B.100 (or L.100) table “Households and Nonprofit Organizations” of the Z.1, or Flow of Funds release. It would seem that the sources as given in the Guide to the Flow of Funds would provide accurate numbers for financial assets. The most obvious caveats for the overall assets total are:

  • B.100 lumps together households and nonprofits. 100a breaks out nonprofits, so one should be able to derive a households-only number for each category, but it is unclear whether in all of the Fed's sources the separation is really explicit, and the 100a table is not documented in the Guide to the Flow of Funds.
  • Current value of residential property is a major component of assets, but the method for calculating current value is not documented (the footnote in the table just says 'at market value').
  • Depreciation, or consumption of residential and durable-goods capital, is bound to be a rather soft number no matter what the assumptions are in calculating it. (This comes from BEA tables A16 / series M1RFI655ONOC and A18 / series M1CTOTL1EQ00)
  • Both the asset and liability totals are derived from a very large number of sources, and there is no documentation or explicit analysis of errors and omissions.
  • The home mortgage debt total is derived quite indirectly and seems to be subject to large possible error; see summary from another page above.

American Housing Survey 2007

2007. Census Bureau report

“American Housing Survey for the United States: 2007”

The AHS asks both what the house actually cost and what the owner thinks it is worth now. For actual cost this is broken down extensively, into closing costs, land, building, etc. Below is some sample data showing how one could derive an estimate of aggregate housing stock value.

AHS overview of methodology

11 Aug 2008. Background doc on Census website.

“How American Housing Survey Data are Collected”

“The AHS Is a Household Survey

The AHS is a household survey that asks questions about the quality of housing in the United States. In gathering information, Census Bureau interviewers visit or telephone the household occupying each housing unit in the sample. For unoccupied units, they obtain information from landlords, rental agents, or neighbors.

The AHS Is Actually Two Surveys

The AHS conducts a national survey and a metropolitan area survey. Both surveys are conducted during a 3- to 7- month period. The national survey, which gathers information on housing throughout the country, interviews at about 55,000 housing units every 2 years, in odd-numbered years.

The metropolitan area survey consists of 47 metropolitan areas, where householders are interviewed every 6 years. We gather data for about 14 metropolitan areas on an even numbered year until all 47 metropolitan areas are surveyed. The cycle begins again 6 years later. Every 4 years, six of the largest metropolitan areas are included with the national sample. Since 1984, each metropolitan area is represented by a sample of at least 3,200 designated housing units. The units are divided between the central city and the rest of the metropolitan area.

A sample of housing units in all survey areas was selected from the decennial census. These are updated by a sample of addresses obtained from building permits (for new construction) to include housing units added since the sample was selected. … The survey goes back to the same housing units on a regular basis, recording changes in characteristics, adding and deleting units when applicable. This cross-sectioning of the housing inventory gives a picture of houses and households as they change over long periods of time.”

Fed method for calculating home value

4 Nov 2008. Personal communication from

[I asked: 'In the Z.1 table B.100, Balance Sheet of Households and Nonprofit Organizations, line 4 (household real estate assets), the footnote to the table only says 'At market value'. Could you please refer me to a publication that describes how this number is calculated?']

“We do not currently have a publication with an explanation of the calculation of this series available to the public. The series shown on table B.100, line 4 is the market value of homes owned by households (as opposed to those owned by businesses or nonprofit organizations). The market value includes mortgage debt on those properties and equity (the latter is calculated as a residual). We benchmark the series to the American Housing Survey. It is calculated from quarter-to-quarter based on (i) adjusting the value of existing homes using OFHEO's purchase-only index of home prices and (ii) net investment in residential real estate from the BEA.

Elizabeth Ball Holmquist Flow of Funds Division of Research and Statistics Federal Reserve Board”

FOF started using Loan Performance instead of OFHEO in Q3 2008

21 Dec 2009. Calculated Risk.

“Flow of Funds: Change in House Price Index. by CalculatedRisk”

“1) Starting with the Q3 2008 report, the Federal Reserve switched from the OFHEO / FHFA house price index to the LoanPerformance house price index. From the Fed in the notes to the Q3 2008 report:

The market value of residential real estate (B.100, B.102, and B.103) has been revised from 2000:Q1 forward to reflect improved data sources. The value of owner-occupied housing in 2001:Q3, 2003:Q2, and 2005:Q2 is now benchmarked to data from the American Housing Survey, and changes in the value of single-family homes in non-benchmark quarters are now estimated using a repeat-sales house-price index from LoanPerformance (a division of First American CoreLogic). Previously we used a price index from the Federal Housing Finance Agency (formerly the Office of Federal Housing Enterprise Oversight).

2) LoanPerformance revised their index in Q3 2009, and this showed up as a substantial change for household real estate value. From the Q3 2009 report:

The market value of residential real estate (B.100, B.102, and B.103) has been revised from 2000:Q1 forward to reflect revised data for the repeat-sales house-price index from LoanPerformance (a division of First American CoreLogic).

The reason this change wasn't obvious in Q3 2008 was that the value of household real estate didn't change significantly for the most recent quarters (although using the LoanPerformance index showed a larger bubble).”

Case-Shiller and Loan Performance HPI are close to parallel

21 Dec 2009. Calculated Risk.

“House Price Indices: Case-Shiller and LoanPerformance. by CalculatedRisk”