"... In PAGE 5: ... Older, more intelligent, and more educated individuals tend to commit less crime, because their skill levels are higher. See Table1 for a breakdown of criminal participation by educational attainment and Table 2 for criminal participation rates by age as reported in the 1980 National Longitudinal Survey of Youth #28NLSY#29. Further, the model suggests #28and the data con#0Crm#29 that di#0Berent types of crime exhibit di#0Berent correlations with age, intelligence, and education depending on the skill requirements of the criminal activity.... In PAGE 11: ...#29 are largely una#0Bected by previous criminal experience or the market skills of criminals. The fact that most street criminals are of low ability #28as measured by IQ #5B18, 42#5D or the Armed Forces Qualifying Test #28AFQT#29 scores discussed below#29, have little education #28 Table1 #29, and are very young #28see Figure 1 and Table 2#29 suggests that the returns to traditional market skills are substantially lower in the criminal sector than the legitimate labor market. We, therefore, begin by studying crimes that do not require skill investments, assuming that N H =0.... In PAGE 26: ...2 Examining the Link Between Education and Crime The model predicts that human capital is negatively correlated with criminal participation. Table1 presents some simple statistics on educational attainment and self-reported criminal activity among 20-23 year old men in the NLSY. Dropouts are much more likely to engage in crime than high school graduates and individuals attending college.... In PAGE 48: ... Significant at 0.10 level. ** Significant at 0.05 level. Table1 0: Effects of HS Graduation and Punishment by Age, Race, and Ability (1980 NLSY) High School Graduation State Punishment Rate Notes: Sample includes all men at ages greater than or equal to 18 in 1980. All regressions include the following regressors: high school graduation status, black and hispanic indicators, whether the individual lived with both his natural parents at age 14, region of current residence, SMSA status, local unemployment rates, and state punishment rate (number of adults incarcerated / number of reported property and violent index crimes).... In PAGE 49: ... Significant at 0.10 level. ** Significant at 0.05 level. Table1 1: Effects of Criminal Status in 1980 on Subsequent Hours Worked, Wage Rates, and Labor Income (1980-93 NLSY) Notes: Sample includes all men at ages greater than or equal to 18 when they were no longer enrolled in school. All regressions include the following regressors: experience, experience-squared, high school graduation status, interactions between high school graduation and experience and experience-squared, black and hispanic indicators, AFQT percentiles, interactions between AFQT and black and hispanic, whether the individual lived with both his natural parents at age 14, region of current residence, SMSA status, local unemployment rates and population levels.... In PAGE 50: ...05. Table1 2: Reductions in Social Costs (1996 Dollars) of Crime (Victim amp; Incarceration Costs) Attributed to High School Graduation (Based on 1980 NLSY Estimated Effects) 1 Per inmate operating expenses in a state prison ($20,100) taken from Stephan, J.... In PAGE 51: ...3858) 1 Punishment measure = (total population incarcerated)/ (total index crime rate from UCR) for each state. Log(Crime Rate): Table1 3: Effects of Education on Crime... ..."

"... In PAGE 17: ... These results therefore strengthen the results reached in Betts (1995). Table2 searches for positive age dependency in another way. In these regressions, the dependent variable is the slope of the log earnings:age profile obtained from 1980 Census data.... In PAGE 17: ... If I find evidence of a positive link, then the impact of additional school spending may manifest itself only as workers age. 14 The results appear in Table2 . First, the table shows that the slope of the predicted log wage:age profile is steeper for more highly educated workers, as indicated in the work by Hanoch (1967).... ..."

"... In PAGE 22: ... A complete description of the data and information collected is given in Appendix B. Table1 contains a comparison of mean values for each of these components across blacks and whites.18 For this group of young black and white males several differentials emerge.... In PAGE 45: ...elow the legislated minimum wage of $3.35*40 hours = $134.00. The means in Table1 are calculated as follows: the unemployment rate is the fraction of respondents in the sample who do not have a job in the first week of April 1985; the mean unemployment duration is the Kaplan-Meier mean of the unemployment durations following the job spells; the fraction of completed spells ending in unemployment is the number of respondents employed in April 1985 that transition to unemployment before December 1988 divided by the total number of respondents employed in April with jobs that end prior to December 1988; the mean earnings is the mean of weekly wages from jobs that are ongoing in April 1985; the mean wage offer is the mean of the weekly wages of jobs that start after April 1985; and the mean job duration is the... ..."

"... In PAGE 13: ... Thus one can view W as a monotonic function of E whose elasticity probably is not unity and need not even be constant. Table2 shows the parameter estimates describing various characterizations of the cases in Section II using the data on job satisfaction for young men ages 26-31 in 1988. I present the results of estimating these ordered probits as the impact of a ten log-point increase in the independent variable on the probabilities of being in each of the three job-satisfaction categories.... In PAGE 14: ... Neither of these models describes job satisfaction in 1988 as well as does a simple version of Case III, in which I try to capture the idea of regret/disappointed expectations. The estimated effects from the ordered probits that are presented in the third row of Table2 are on annual earnings in 1988 and on 78W^i,88, what the average earnings in 1988 of a worker with individual i s observable characteristics would have been had the distribution of returns to someone with those characteristics remained unchanged between 1978 and 1988. This equation is thus essentially identical to (2c), except actual earnings replace full earnings.... In PAGE 15: ... While the data from these two cross-sections of young men do not allow distinguishing between Cases III and IV, they can be used to examine the relative importance of regrets about the returns to observable and unobservable characteristics in the determination of job satisfaction. The remaining effects presented in Table2 are from ordered probits that include as independent variables combinations of measures that decompose earnings shocks into those to the returns to the observable and unobservable components of skills. The first term is [u88 - u^p,78], the residual from an equation describing earnings in 1988 minus what that residual would have been at the p th percentile of the earnings distribution if the distribution of observed earnings had remained unchanged since 1978.... In PAGE 15: ...) relative to what could have been expected in 1978. The fourth and fifth rows of Table2 present the results of including these two measures... In PAGE 16: ..., that are not determinants of the education and job tenure that they accumulate); or the returns to these unobservables simply do not affect how they perceive their jobs.7 The final equation in Table2 includes an interaction of [u88 - u^p,78] and [W^88 - 78W^88] to test whether unusually high (low) unexpected returns to both observables and unobservables interact to generate unusually high (low) job satisfaction. With the inclusion of this interaction term all three coefficients are significantly nonzero; and the results suggest that unusual returns along both observable and unobservable dimensions affected job satisfaction in 1988.... In PAGE 22: ... In each equation I also include indicator variables for JS = 4 and JS = 3 at time t-6. The results are presented as in Table2 : The first two numbers are the impacts of ten log-point increases in the independent variables on the probabilities of responding JSt =4 or JSt = 3, while the t-statistic on the ordered-probit coefficient is in parentheses below these derivatives. The first thing to note is that the hypotheses implicit in Cases I and II both make some sense in these estimates that account for individual heterogeneity.... ..."

"... In PAGE 21: ... One can make a strong case that gender and race are exogenous, but sibship size (number of siblings), religion, and parental schooling can be considered endogenous to the family. Results for these models are presented in Table3 . Like previous research, our OLS and probit cross-section Models (A) and (B) show that years with a single parent or stepparent have negative and significant effects on schooling outcomes.... In PAGE 21: ... Years with a stepparent have negative and significant effects on college attendance and graduation.9 The fixed effects estimates in Table3 tell a different story. For all four schooling outcomes, the family structure variables are not statistically significant.... In PAGE 21: ... As reported in Tables 1 and 2, more of the variation in schooling outcomes occurs between than within households. 9 We have experimented with alternative specifications in Table3 and found our results to be robust. We used dummy variables for family structure instead of years living in a particular family structure.... In PAGE 21: ... We used dummy variables for family structure instead of years living in a particular family structure. The estimates presented in Table3 fit the data better than those using family structure dummies but tell the same story. In Model (B) we substituted dummy variables for number of siblings to account for nonlinearities in the effect of family size on outcomes.... ..."

"... In PAGE 12: ...Descriptive statistics are shown in Table1 . Table 2 gives more detail on the distribution of training by showing selected percentiles of the positive distribution of both the stock of training and of training during the previous year.... In PAGE 35: ... Recall that one of the control variables in our wage growth regression is the logarithm of the number of weeks it takes a new employee in the most recently filled position to become fully trained and qualified if he or she has the necessary school provided training but no experience in the job, which we refer to as quot;job complexity quot;. Consistent with this interpretation, quot;job complexity quot; is positively related to wage growth, as can be seen in column 1 of Table1 2, which reports the key coefficients on training and job complexity in our preferred EOPP specification (i.e.... In PAGE 46: ... Table1 , continued Variable Mean Std. Dev.... In PAGE 66: ... Results are shown in Appendix Tables 1 and 2. Table1 shows the results for the four-period setup. There appears to be a downward bias in each specification due to the spline functional form, in most cases small.... In PAGE 68: ...Appendix Table1 , continued % bias at median True functional form Probability of forgetting No duration error Duration error present T.75 0.... ..."

"... In PAGE 21: ...69% (the number of moves per person-year in the data). Table5 shows the predictions from this model: about 73% of the people never move, and of those who do move, about 15% move more than once.22 The NLSY data are quite different: about 82% never move, and about 53% of movers move more than once.... ..."

"... In PAGE 59: ...V.B.1. Introduction In the National Longitudinal Survey of Youth (NLSY) data, the African American abortion rate is less than a third of the white abortion rate ( Table13 ).38 However, Lundberg and Plotnick (1995) provide evidence that both African Americans and whites underreport the actual number of abortions.... ..."