Arthur B. Kennickell

Currents and Undercurrents: Changes in the Distribution of Wealth, 1989-2004

This paper considers changes in the distribution of the wealth of U.S. families over the 1989-2004 period using data from the Survey of Consumer Finances (SCF). Real net worth grew broadly over this period. At the same time, there are indications that wealth became more concentrated, but the result does not hold unambiguously across a set of plausible measures. For example, the Gini coefficient shows significant increases in the concentration of wealth from 1989 to 2004, but the wealth share of the wealthiest one percent of families did not change significantly. Graphical analysis suggests that there was a shift in favor of the top of the distribution, while for the broad middle of the distribution increases were about in proportion to earlier wealth. Within this period, there are other interesting patterns. For example, from 1992 to 2004 the wealth share of the least wealthy half of the population fell significantly to 2.5 percent of total wealth. The data show little in the way of significant distributional shifts since the 2001 survey. The paper also presents some information on underlying factors that may explain a part of the distribution of wealth, including capital gains, saving behavior and income, inheritances, and other factors. There are two special topic sections in the paper. The first presents information on the distributions of wealth of African American and Hispanic families. The second presents information on the use of debt across the distribution of wealth.

The Importance of Business Owners in Assessing the Size of Precautionary Savings

Not properly accounting for differences between business owners and nonbusiness owners in studies of household wealth can lead to erroneous conclusions about the significance of different saving motives. Using data from the Panel Study of Income Dynamics from the 1980s and 1990s, we show that within samples of both business owners and nonbusiness owners, the amount of precautionary savings with respect to labor income risk is modest and accounts for less than 10 of total household wealth. Previous large estimates of the size of precautionary balances resulted from pooling these two groups together. Such pooling is inappropriate given that business owners face higher labor risk and accumulate more wealth than nonbusiness owners for reasons unrelated to precautionary motives.

Pensions, Social Security, and the Distribution of Wealth

This paper uses the Survey of Consumer Finances (SCF) to examine pension coverage, estimate Social Security and pension wealth for U.S. households in 1989 and 1992, and estimate the effects of pension wealth on non-pension net worth. As expected, the SCF data show that including pensions and Social Security in net worth makes the distribution more even. The analysis of the effects of pension wealth on other types of savings indicates that there is a negative effect of defined benefit plan coverage on non-pension net worth. Surprisingly, the effect of defined contribution plans, such as 401(k) plans is insignificant.

Modeling wealth with multiple observations of income: Redesign of the sample for the 2001 Survey of Consumer Finances

This paper reports on research that underlies the redesign of a key part of the sample for the Survey of Consumer Finances (SCF) for the 2001 wave of the survey, and it documents the implementation of the new design. This work builds on the history of sampling research for the survey dating back to the 1983 wave. The sample for the SCF includes an oversample of relatively wealthy families. The stratified design is based on a mapping from observations of components of income to an estimate of wealth. The mapping is imperfect, and three imperfections seem particularly important: First, at any given time, rates of return that connect assets with capital income may vary widely across individuals depending on luck, information, and local economic conditions. Second, some assets, such as 401(k) accounts, do not generate regularly observable returns. Finally, transitory factors, such as the timing of income receipts or unusually good luck, may cause the income that is observed in a given period to have a noisy relationship to the underlying assets that generate those returns. It is very difficult to address the first two concerns directly. However, the third can be addressed by using multiple observations of income to model wealth; an effort to do so is the principal focus here. The first section of this paper gives a brief overview of the SCF and the sample design. The next section provides motivation for a reexamination the sample design. The third section describes the implementation of the redesign of the sample for the 2001 SCF. A final section discusses areas for future research. I. Overview of the SCF and Sample design The SCF is undertaken every three years by the Federal Reserve Board (FRB) primarily as a study of household wealth and use of financial services. To this end, the subject matter of the questionnaire focuses on detailed components of assets and liabilities and supporting information. Because household wealth in the U.S. is highly concentrated (Kennickell, 2000b), a very large sample would be required to make reliable wealth estimates without some type of oversampling by wealth. At the same time, the problem of adequate representation of wealth is amplified by the nonrandom nature of nonresponse to the survey. Although many factors enter into nonresponse in the SCF (Kennickell, 1999b), it is clear that wealthier respondents are less likely to participate. Lower participation may be driven by a greater sensitivity about privacy issues, a greater perceived value of the time required for the interview, the greater difficulty interviewers face in getting beyond gatekeepers in order to request participation, or other factors. The net result is that without some means of addressing the nonrandom nonresponse, many of the survey estimates would be severely biased. The SCF sample is a dual-frame design. One part is a multi-stage national area-probability design (Tourangeau et al., 1993), which provides good coverage of the general population. This sample is selected by the National Opinion Research Center at the University of Chicago, the contractor for data collection since 1992. The second part is a list sample, which is selected by FRB staff from statistical records derived from tax returns. These records are provided by the Statistics of Income Division (SOI) of the Internal Revenue Service to the FRB under strict controls on the use of the data. This file is generated from a sample from the full set of tax returns filed, and it is specially edited using procedures designed to yield data to support tax research at the Office of Tax Analysis and the Joint Economic Committee (see IRS, 1998). Since the 1989 survey, the SOI data used for the SCF sample have come from the tax year (generally, the year in which the reported income was earned) two years before the date of the survey. The list sample is stratified using a “wealth index” computed using income data in order to predict a rank ordering of people by wealth. This stratification allows both oversampling of wealthy families and targeted nonresponse adjustment. The area-probability and list samples are pooled using weights that are designed to respect the relative strengths of each part of the sample (Kennickell and Woodburn, 1999). Because the wealth index used in the list sample stratification has such a powerful effect on the estimation efficiency of the SCF, refinement of the index has always been an important part of the methodological research supporting the survey (see Kennickell 1999a for a history). The list sample for the 1983 survey, the first of the series, was stratified by income categories created using a set of rules that have not been cleared for public release. The 1989 SCF was the first to use a wealth index approach, and the construction of that index is discussed in Heeringa et al. (1994). In its pure form, this general type of index for a given time might be expressed as

Precautionary Savings and the Importance of Business Owners

In this paper, we show the pivotal role business owners play in estimating the importance of the precautionary saving motive. Since business owners hold larger amounts of wealth than other households for non-precautionary reasons and also face highly volatile income, they induce a correlation between wealth and income risk regardless of whether or not a precautionary saving motive exists. Using data from the Panel Study of Income Dynamics in the 1980s and the 1990s, we show that among both business owners and non-business owners, the size of precautionary savings with respect to labor income risk is modest and accounts for less than ten percent of total household wealth. However, pooling together the two groups leads to an artificially high estimate of the importance of precautionary savings. New data from the Survey of Consumer Finances further confirms that precautionary savings account for less than ten percent of total wealth for both business owners and non-business owners. Thus, while a precautionary saving motive exists and affects all households, it does not give rise to high amounts of wealth in the economy, particularly among those households who face the most volatile stream of income.

Measuring the impact of data protection techniques on data utility: evidence from the survey of consumer finances

Despite the fact that much empirical economic research is based on public-use data files, the debate on the impact of disclosure protection on data quality has largely been conducted among statisticians and computer scientists. Remarkably, economists have shown very little interest in this subject, which has potentially profound implications for research. Without input from such subject-matter experts, statistical agencies may make decisions that unnecessarily obstruct analysis. This paper examines the impact of the application of disclosure protection techniques on a survey that is heavily used by both economists and policy-makers: the Survey of Consumer Finances. It evaluates the ability of different approaches to convey information about changes in data utility to subject matter experts.

Darkness Made Visible: Field Management and Nonresponse in the 2004 SCF

Nonresponse in field surveys is the joint outcome of the decision of survey staff to apply effort to inform and persuade respondents, and the evaluation of such inputs by respondents. In most such surveys, the field staff are under great pressure to produce completed interviews. Thus, as discussed in Kennickell [2004], they have an incentive to apply effort to cases that are most likely, in their view, to be completed with least effort. To the extent that interviewers’ perceptions are unbiased, such behavior would tend to amplify latent patterns of nonresponse. When the characteristics of respondents that affect the likelihood of participation are correlated with variables of analytical interest in the survey, bias results, unless a means can be found of discovering and adjusting for the underlying behavioral structures. But, absent constraints on the behavior of interviewers, the observed outcomes are contaminated by the endogeneity of effort, and only strong a priori assumptions could disentangle the interviewer effects from the respondent effects. To address the problem of endogenous effort, the 2004 Survey of Consumer Finances introduced a phased plan of sample management to make effort more nearly exogenous through the first two of three phases of field work. Thus, nonresponse in these controlled stages should largely reflect respondent characteristics, not a mixture of respondent and interviewer characteristics. The dual frame design of the SCF offers two classes of sample cases for modeling nonresponse. For the area-probability sample, tract-level data are available from the 2000 Census of Population. For the list sample, frame case-specific data based on statistical records derived from tax returns are available. For both set of cases, some interviewer observations are also available. This paper presents estimates of nonresponse models based on these data. Nonresponse in field surveys is the joint outcome of the decision of survey staff to apply effort to inform and persuade respondents, and the evaluation of such inputs by respondents. In most such surveys, the field staff are under great pressure to produce completed interviews. Thus, as discussed in Kennickell [2004], they have an incentive to apply effort to cases that are most likely, in their view, to be completed with least effort. To the extent that interviewers’ perceptions are unbiased, such behavior would tend to amplify latent patterns of nonresponse. When the characteristics of respondents that affect the likelihood of participation are correlated with variables of analytical interest in the survey, bias results, unless a means can be found of discovering and adjusting for the underlying behavioral structures. But, absent constraints on the behavior of interviewers, the observed outcomes are contaminated by the endogeneity of effort, and only strong a priori assumptions could disentangle the interviewer effects from the respondent effects. To address the problem of endogenous effort, the 2004 Survey of Consumer Finances (SCF) introduced a phased plan of sample management intended to make effort more nearly exogenous through the first two of three phases of field work. As described in more detail later in this paper, interviewers were given a flexible protocol intended to ensure that all sample cases were exposed to a certain level of effort and that two specific points in the course of the application of effort were marked for cases that had not been completed earlier. Those points classify the field operations on the cases into at most three phases, the last of which is limited only by the close of field work. If the framework holds, then response within the first two phases may be taken as independent of behavioral variations in the level of effort applied. The first section of the paper provides an overview of the SCF and the approach to management of field resources developed for the 2004 survey and describes how this protocol worked in practice. The second section presents models of nonresponse in different phases of the field effort, conditional on both tract-level and case-level data. The final section summarizes the key findings and points toward the next step for the SCF.

Using range techniques with CAPI in the 1995 Survey of Consumer Finances

from changes in sample composition over this Table 1: Item response rates over time, AP Sample, Unweighted House Chckng Stocks Income Have item? Yes 1989 55.2 80.5 16.6 100.

Getting to the Top: Reaching Wealthy Respondents in the SCF

This paper examines the effort devoted to securing interviews with a very wealthy part of the sample for the 2007 Survey of Consumer Finances (SCF). Only about a quarter of the group completed an interview. At the close of the field period, more than a third of this part of the sample was judged by the field staff to be still workable—that is, those cases were neither complete nor final refusals. The evolution of the field work was driven both by the behavior of respondents and the behavior of the field staff. The paper uses the formal data coded in the call records for each case to describe the work. But that information is inconclusive about the factors that drove the work. However, informal notes in the call records do provide a clear picture of the points of resistance among respondents. Although it was difficult to locate, contact, and convince respondents of the legitimacy and value of the survey, it appears that the ultimate constraint in a large proportion of cases was the length of the interview—potentially several hours for these respondents. Examination of the available auxiliary data provides little evidence of nonresponse bias.

Try, Try Again: Response and Nonresponse in the 2009 SCF Panel

The 2007 SCF was designed as a continuation of a series of cross-sectional surveys on the financial condition of U.S. households. In light of the serious economic downturn that followed that survey, the Federal Reserve Board decided to pursue a second interview with the survey participants to understand how the aggregate changes played out across households. Great care was taken to prepare interviewers to deal with respondents, who would not have expected an additional contact and some of whom had earlier expressed a strong desire never to be bothered again. Ultimately, the survey achieved a re-interview rate of almost 89 percent and a relatively low item nonresponse rates for such a complex survey. This paper uses the formal and informal paradata to examine key factors in survey response. If the nonrespondents to the re-interview are representative of marginal respondents in both years, there is an advantage in studying the group, because so much is known about them from their earlier interview and the process of obtaining that interview.