This paper documents and attempts to explain the observed disparities
between unemployment rates computed from contemporaneous and retrospective CPS
data. The maintained hypothesis is that the discrepancies are consistent with
different definitions of unemployment between the two measures. The longitudinal
nature of the CPS, which allows a respondent's answers to be matched between one
year and the next, is exploited to examine two commonly expressed shortcomings
in the contemporaneous definition. I find that relative to the retrospective
measure, more workers with weak labor force attachment are considered unemployed
in the contemporaneous rate. In addition, discouraged workers, who are
classified as out of the labor force according to the contemporaneous definition,
may be counted as unemployed in the retrospective.
Cross section estimates of the union wage effect are typically much
larger than estimates derived from within estimators using panel data. Two
competing explanations for this difference have been advanced. The first is
that the cross section estimates suffer from an omitted variables bias due to
a correlation between unobserved productivity and union status which biases
the cross section estimator upwards. The second is that measurement error
in union status is more severe in the changes than in the levels, imparting a
more severe downward bias to the panel estimator.
This paper derives a method of moments estimator which allows for
both effects, nested within the same model. The binary nature of the
explanatory variable is exploited to derive an estimating model which allows
simultaneous estimation of both the structural parameters of the model and
the parameters of the measurement error process. When the estimator is
applied to sample of men from the PSID we find that allowing for measurement
error does lead to a larger estimate of the union wage effect than the usual
within estimator, but that most of the difference between the cross section
and the panel estimates is not due to measurement error in the union
variable. Further, the estimates of the extent of measurement error are
close to those found in a validation study of the PSID.
In this paper we use data on brothers, and fathers and sons, to estimate the economic
returns to schooling. Our goal is to determine whether the correlation between earnings and
schooling is due, in part, to the correlation between family backgrounds and schooling. The
basic idea is to contrast the differences between the schooling of brothers, and fathers and sons,
with the differences in their respective earnings. Since individuals linked by family affiliation
are more likely to have similar innate ability and family backgrounds than randomly selected
individuals our procedure provides a straightforward control for unobserved family attributes.
Our empirical results indicate that in the sample of brothers the ordinary least squares
estimates of the return to schooling may be biased upward by some 25% by the omission of
family background factors. Adjustments for measurement error, however, imply that the
intrafamily estimate of the returns to schooling is biased downward by about 25% also, so that
the ordinary least squares estimate suffers from very little overall bias. Using data on fathers
and sons introduces some ambiguity into these ﬁndings, as commonly used speciﬁcation tests
reject our simplest models of the role of family background in the determination of earnings.
This paper examines the properties and prevalence of measurement
error in longitudinal earnings data. The analysis compares Current
Population Survey data to administrative Social Security payroll tax
records for a sample of heads of households over two years. In contrast
to the typically assumed properties of measurement error, the results
indicate that errors are serially correlated over two years and
negatively correlated with true earnings (i.e., mean reverting).
Moreover, reported earnings are more reliable for females than males.
Overall, the ratio of the variance of the signal to the total variance is
.82 for men and .92 for women. These ratios fall to .65 and .81 when the
data are specified in first-differences. The estimates suggest that
longitudinal earnings data may be more reliable than previously believed.
This paper uses a new survey to contrast the wages of genetically
identical twins with different schooling levels. Multiple measurements of
schooling levels were also collected to assess the effect of reporting
error on the estimated economic returns to schooling. The data indicate
that omitted ability variables do not bias the estimated return to
schooling upward, but that measurement error does bias it downward.
Adjustment for measurement error indicates that an additional year of
schooling increases wages by 12-l6t, a higher estimate of the economic
returns to schooling than has been previously found.
We propose a general method of moments technique to identify measurement error in self-reported
and transcript-reported schooling using differences in wages, test scores and other covariates to discern the
relative verity of each measure. We also explore the implications of such reporting errors for both OLS and
IV estimates of the returns to schooling. The results cast a new light on two common findings in the
extensive literature on the retums to schooling: “sheepskin effects” and the recent IV estimates, relying on
“natural experiments” to identify the payoff to schooling. First, respondents tend to self-report degree
attainment much more accurately than they report educational attainment not corresponding with degree
attainment. For instance, we estimate that more than 90 percent of those with associate’s or bachelor’s
degrees accurately report degree attainment, while only slightly over half of those with l or 2 years of college
credits accurately report their educational attainment. As a result, OLS estimates tend to understate returns
per year of schooling and overstate degree effects. Second, because the measurement error in educational
attainment is non-classical, IV estimates also tend to be biased, although the magnitude of the bias depends
upon the nature of the measurement error in the region of educational attainment affected by the instrument.
This paper tests the rational expectations lifecycle model of consumption
against (i) a simple Keynesian model and (ii) the rational expectations
lifecycle model with imperfect capital markets. The tests are based upon the
relative responsiveness of consumption to income changes which can be
predicted from past information and income changes which cannot be predicted.
Problems caused by measurement error in the income changes are circumvented by
using the innovations from a vector autoregression of the measures of the
determinants of income to form a noisy instrument for the unanticipated change
in incomﬁ and using the lagged values of the measures of the income
determinants to form an instrument for the anticipated income change. We show
that the Keynesian model implies that the regression coefficients relating the
change in consumption to the instruments for the anticipated and unanticipated
components of the income change should be equal. The lifecycle model (with
perfect capital markets) implies that only the instrument for the
unanticipated component should affect consumption. The empirical results
support the lifecycle model. In addition, we incorporate capital market
imperfections into our empirical formulation of the lifecycle model by
assuming that the marginal interest rate at a point in time is a
differentiable, concave function of net assets. This leads to a test for
capital market imperfections based upon whether consumption responds
differently to positive and negative predictable changes in income. Our
results are inconclusive.
Both marital status and computer usage on the job have been found to increase earnings by as much as two
additional years of schooling. If correct, these ﬁndings suggest that factors other than long-term human
capital investments are key determinants of earnings. Data on identical twins are used in this paper to sweep
out selection effects and examine the effect of marital status and computer usage on wages. Within-twin
estimates indicate that, unlike education, job tenure and union status, neither marital status nor computer
usage have a large or signiﬁcant effect on wages.