Macro Roundup

“….It appears that cognitive skills during adolescence, as measured by scores on the Armed Forces Qualification Test (AFQT), are strongly associated with these gaps. For example, conditional on having the median AFQT score, the racial gaps in both upward and downward mobility are relatively small. Consistent with previous studies linking AFQT scores to racial differences in adult outcomes (for example, Neal and Johnson, 1996; Cameron and Heckman, 2001), I do not interpret these scores as measuring innate endowments but rather as reflecting the accumulated differences in family background and other influences that are manifested in test scores.6 If these results are given a causal interpretation, they suggest that actions that reduce the racial gap in test scores could also reduce the racial gap in intergenerational mobility…. The effects of including one’s AFQT score on rates of upward mobility are shown in panel C of figure 8. Here, the results provide a relatively clean and compelling story. For both blacks and whites, upward mobility rises with AFQT scores in a fairly similar fashion. There are especially sharp gains in upward mobility associated with increases in test scores at the low end of the AFQT distribution. Upward mobility continues to rise at a somewhat slower but still strong rate in the middle and in the upper half of the AFQT distribution. Remarkably, the lines for blacks and whites are relatively close throughout the AFQT distribution. For example, the black-white gap in moving out of the bottom quintile is only 5.2 percentage points for those with median AFQT scores, compared with the unconditional gap of 27 percentage points. This suggests that cognitive skills measured at adolescence can “account” for much of the black-white difference in upward mobility. This result echoes previous findings by Neal and Johnson (1996) and Cameron and Heckman (2001), who have also found that AFQT scores can account for much of the racial gap in adult earnings and college enrollment rates. As with these aforementioned studies, I interpret this finding as reflecting the cumulative effect of a broad range of family background influences rather than reflecting only innate differences.28 The effects of AFQT scores on downward mobility (panel D of figure 8) are also quite striking. The lines for whites and blacks converge quite a bit and for a broad swath of the AFQT distribution, the racial gap is below 10 percentage points and is not statistically different from 0. Therefore, as was the case with upward mobility, test scores during adolescence are strongly associated with rates of downward mobility….”

“…Our most striking finding, however, is that when we (nonparametrically) control for AFQT scores, most of the racial gap in IGM disappears—whether measured via transition probability or upward mobility. We are careful in our interpretation of this finding. First, existing studies suggest that these scores largely reflect accumulated differences in family background and other influences, and are not primarily a measure of innate endowments; indeed, there is evidence that black-white differences in tests scores can be strongly affected by environmental influences….Second, like most of the existing IGM literature, our analysis is essentially descriptive in nature and one should be cautious in attaching causal interpretations to them. Third, all measures used in the present paper are based on relative positions and do not pertain to movements out of absolute thresholds such as from under the poverty line. However, our finding that there is no racial gap in mobility, conditional on cognitive test scores in adolescence, is new and merits further analysis to establish possible policy implications, such as the importance of early-life interventions relative to, say, access to affirmative action later in life…. We estimate the effect of AFQT scores on upward transition probabilities separately by race. Our dependent variable is the probability of leaving the bottom quintile. Figure 6 shows the result of this exercise. We find that conditional on AFQT scores, whites have only slightly higher likelihood of exiting the bottom quintile and that this gap does not vary a great deal across the AFQT distribution. For example, at the 25th interval of our normalized AFQT scores, the transition probability for whites is 0.63 and for blacks is 0.61, or a difference of just 2 percentage points. At the 10th interval, the gap is about 7 percentage points and at the 75th interval, the gap is about 15 percentage points. At no point in the AFQT distribution can we reject the hypothesis that the transition probabilities are the same….”

Debopam Bhattacharya and Bhashkar Mazumder, " A nonparametric analysis of black-white differences in intergenerational income mobility in the United States,"Quantitative Economics, 2011, https://onlinelibrary.wiley.com/doi/epdf/10.3982/QE69

Moffitt took a look at the research looking at male earnings volatility and attempted to adjudicate it, finds little change in male earnings volatility in the past 35 years

Note attached copy is from a version of the paper from July 2020

"In this paper we compare estimates of earnings volatility from SIPP survey data and the SIPP GSF, which links SIPP survey respondents to their respective administrative earnings records. We show that the trends in the SIPP survey and the SIPP GSF are similar. Volatility increases modestly in the SIPP GSF and declines modestly in the SIPP survey data between the mid-1980s and the early 2010s. Levels in the SIPP GSF are considerably higher than in the SIPP survey data. The results are robust to age-adjusting, more aggressively trimming low earnings, the use of imputations, and different methods of handling population representativeness. We also show that differences in the underlying earnings distribution explain the differences in the trend and level of volatility in the SIPP GSF and SIPP survey data. When we force the earnings distributions in the two datasets to be the same, we find that the differences in levels of volatility largely disappear, and the minor differences in trends that exist are reduced. This paper is part of a larger project designed to consider estimates of earnings volatility across multiple datasets. As part of this project, we also compare volatility in the SIPP GSF and SIPP survey data to that in PSID. We compare volatility using the same sample selection criteria and methods and find that there remain differences in both the level and trend in volatility. Volatility in the SIPP GSF is higher than the PSID, while volatility in the SIPP survey is lower, and both shower smaller relative increases between 1999 and 2010. Even when the earnings distributions are forced to be the same across the three datasets, there remain differences in both level and trend in volatility between the SIPP and the PSID. We find that, for both the SIPP GSF and the SIPP, survey volatility falls to a lower level than the PSID when weighting the cross-sectional earning distribution to the PSID, implying that volatility is higher in the PSID at a fixed level of inequality. More broadly, the results of this paper and the other papers in this group have important implications for the broader literature on volatility. Using 3 different survey data sets and 3 different administrative data sets, we find broad agreement that volatility has shown no major trend upward or downward in the past 35 years. Thus our paper, along with the others, finds no difference in the evidence obtained from survey and administration data. Another finding that has broad applicability is that, in some cases, differences in the underlying cross-sectional earnings distribution in different data sets plays a role in explaining differences in both levels and trends in volatility. The substantial differences across data sources in the level of inequality, across methods used to estimate volatility in the weight placed on different portions of the earnings distribution, and across sample definitions that may themselves alter the level and trend in inequality, could contribute to some of the disagreement that has developed in this literature...."

Kevin Lang and Michael Manove, “Education and Labor Market Discrimination,” American Economic Review, June 2011, https://www.nber.org/papers/w12257

“…First, they maintain that we should examine black white wage differentials without conditioning on education, because education is endogenous. Their argument would be much more compelling if blacks obtained less education than equivalent whites. In that case, we might argue that blacks get less education because they expect to face discrimination in the labor market, so that controlling for education understates the importance of discrimination. If blacks obtain more education, however, because they anticipate labor market discrimination, failing to control for education understates the impact of discrimination. Consider the following example: the market discriminates against blacks by paying them exactly what it would pay otherwise equivalent whites with exactly one less year of education. Then, to a first approximation, all blacks will get one year more education than otherwise equivalent whites. Controlling only for ability, blacks and whites will have the same earnings, but controlling for education as well as ability, blacks will earn less than whites by an amount equal to the return to one year of education. Even if the higher educational attainment among blacks reflects premarket factors, it may still be appropriate to control for education when measuring discrimination in the labor market. After all, we would still expect the labor market to compensate blacks for their additional education…Blacks get more education than whites with the same AFQT do. So absent discrimination, we would expect blacks to be rewarded for their greater education. The similar earnings of blacks and whites when we control only for AFQT suggest that blacks are not so rewarded. We show that when we control for both education and AFQT, wage differentials between blacks and whites are substantially larger than when we control for AFQT alone…. In our empirical analysis, we proxy ability by performance on the Armed Forces Qualification Test (AFQT). In many cases, we limit the sample to those too young to have completed school as of the time of testing. We find that educational attainment, conditional on AFQT, is higher among blacks than it is among whites, a result that is consistent with our signaling hypothesis. We also predict that whites and blacks will have similar earnings at very low and very high levels of education (not controlling for ability) but that blacks will have lower earnings at intermediate levels of education.We do not test this for women because of selection issues (Derek Neal 2004), but we do confirm the hypothesis for men. Our particular formalization implies that blacks would have levels of education similar to those of whites at low and high levels of ability and levels exceeding those of whites at intermediate levels of ability. This is confirmed only for men (black women have more education than whites at all ability levels but the lowest)….. In our view, the results in this paper cast doubt on an emerging consensus that the origins of the black-white wage differential lie in premarket rather than labor-market factors. Blacks earn noticeably less than whites with the same education and cognitive score. Controlling for measures of school quality (both inputs and composition) leaves this result unaltered. Although it is still possible that the earnings gap reflects differences between blacks and whites that are not captured by our controls, our results should lead us to question whether the gap is solely due to premarket factors and ask whether at least some of the black-white wage differential reflects differential treatment in the labor market….”

Little factoid, at highest level of educational attainment (20 years) black men earned more than their white counterparts (Figure 6)

Paper argues (contrary to Neal and Johnson) that you have to control for education in addition to AFQT scores as black with the same ~ score as whites get more education, but earn less, which is suggestive of labor market discrimination, not premarket characteristics

Derek Neal and William Johnson, “The Role Of Premarket Factors In Black-White Wage Differences,” The Journal of Political Economy, October 1996, https://www.nber.org/papers/w5124

They view education as a“noisy measure of workers skill because it measures an input, not an outcome”which is a good turn of phrase

This paper from 1996 looks at the covariance issue and finds that the AFQT is the variable that matters“we find that this one test score explains all of the black-white wage gap for young women and much of the gap for young men. For today’s young adults, the black white wage gap primarily reflects a skill gap”

“…Recent literature on the black-white wage gap continues to show that a large residual due to race remains, and during the 1980s it grew in size. One interpretation is that the residual gap measures the existence of labor market discrimination. Another interpretation is that imbedded in the residual gap are racial differences in unobservable skills that grew during the 1980s, or that racial differences in these skills remained constant, but their returns grew. To account for these unobservable skills, researchers use the National Longitudinal Survey of Youth (NLSY). The NLSY contains the Armed Forces Qualification Test (AFQT), a direct measure of skills obtained via family and school environments. When researchers include the AFQT composite score, the black-white wage gapnarrows. Doing this, they assume that the AFQT score is a racially unbiased predictor of wages. We present evidence that generates doubts that the AFQT score is racially unbiased. We first show that F-Tests reject the hypothesis that AFQT scores equally predict African American and white wages. Second, we show that a significant difference exists in the ability of variables representing family background, school quality, and psychological characteristics to estimate black and white test scores. Third, we present our estimates of the black-white wage gap using a racially unbiased AFQT composite score to control for racial differences in job skills. The score has a significant independent effect on wages, and a very small effect on the race coefficient.…Our empirical results, which demonstrate that the composite test scores of civilian African Americans in the NLSY, as a predictor of their wages, are racially biased. To do this, we show that F-Tests reject the hypothesis that AFQT composite scores equally predict African American and white wages. Further, when the components of the AFQT composite are used to predict wages, instead of the composite score, the coefficients on the verbal components are positive and significant for African Americans, while the coefficients on the math component is basically zero. The relationship for whites are exactly the opposite. This suggests that the NLSY test scores are measuring different job skills of African Americans and whites….The unconditional black-white gap in wages is about 30% for young men. Those who use the “raw” AFQT score purport to explain approximately two-thirds of the gap. Our “raw” AFQT score explains one-half of the female gap and one-fifth of the male gap. Yet when we use an estimator that “purges” the score of its racial bias, only 12 percent of the female gap is explained, and 5 percent of the male gap is explained….”

William Rodgers and William Spriggs, “What Does The AFQT Really Measure: Race, Wages, Schooling And The AFQT Score,” The Review Of Black Political Economic, June 1996, https://link.springer.com/article/10.1007/BF02690041