Much research (surveyed in Katz and Autor, 1999) documents a substantial widening of the U.S. wage structure since the late 1970s, driven by increases in educational wage differentials and residual wage inequality. The growth in wage inequality was most rapid during the 1980s, and involved a spreading out of the entire wage distribution. Rapid secular growth in the demand for skills, partly from skill-biased technical change (SBTC), combined with a slowdown in the growth of the relative supplyof college workers helps explain these wage changes. Eroding labor market institutions-the minimum wage and unions-further contributed to rising wage inequality.Recent work emphasizes a slowing of wage inequality growth over the last 15 years (David Card and John DiNardo, 2002; Thomas Lemieux, forthcoming). This "revisionist" literature views the 1980s surge in wage inequality as an "episodic" event caused by institutional forces and argues that "modest" inequality growth in the 1990s is inconsistent with a key role for SBTC.We reconsider this revisionist view, focusing on a marked change in the evolution of the U.S. wage structure over the past 15 years and divergent trends in upper- and lower-tail wage inequality. We first document that wage inequality in the top half of the distribution has exhibited an unchecked secular rise for 25 years, but it has ceased growing since the late 1980s (and for some measures narrowed) in the bottom half of the distribution. We next demonstrate that employment growth differed sharply in the 1990s versus the 1980s, with more rapid growth of employment in jobs at the bottom and top relative to the middle of the skill distribution. Borrowing terminology from Maarten Goosand Alan Manning (2003), we characterize this pattern as a "polarization" of the U.S. labor market, with employment polarizing into highwage and low-wage jobs at the expense of middle-skill jobs. We then show how a model of computerization in which computers complement nonroutine cognitive tasks, substitute for routine tasks, and have little impact on nonroutine manual tasks, can rationalize this polarization pattern.

The Ben-Porath (1967) mechanism suggests that prolonging the period during which individuals may receive returns on their investment spurs investment in human capital and causes growth. An important, albeit implicit implication of this mechanism is that the total labour input over a lifetime must increase as longevity does. Otherwise, the incentive to invest in education would not increase. We propose an empirical evaluation of the relevance of this mechanism to the transition from 'stagnation' to 'growth' in todays developed economies. Specifically, we estimate the expected total lifetime working hours of consecutive cohorts of American men born between 1840 and 1970. Our results show that despite a gain of more than 15 years in life expectancy at the age 5, the expected total lifetime working hours have declined by more than 20 percent between the oldest and youngest cohorts. Furthermore, the similarity in the trends and the magnitudes of the determinants of total lifetime labour input between the US and many European countries suggest that our result is not confined to the US experience; rather, it is a robust feature of the process of development. We conclude that the Ben-Porath mechanism has had no effect on the accumulation of human capital during the growth process of the nineteenth and twentieth centuries.

This paper asks the following question: what was the effect of surging immigration on average and individual wages of U.S.-born workers during the period 1990-2004? We emphasize the need for a general equilibrium approach to analyze this problem. The impact of immigrants on wages of U.S.-born workers can be evaluated only by accounting carefully for labor market and capital market interactions in production. Using such a general equilibrium approach we estimate that immigrants are imperfect substitutes for U.S.- born workers within the same education-experience-gender group (because they choose different occupations and have different skills). Moreover, accounting for a reasonable speed of adjustment of physical capital we show that most of the wage effects of immigration accrue to native workers within a decade. These two facts imply a positive and significant effect of the 1990-2004 immigration on the average wage of U.S.-born workers overall, both in the short run and in the long run. This positive effect results from averaging a positive effect on wages of U.S.-born workers with at least a high school degree and a small negative effect on wages of U.S.-born workers with no high school degree.

Revisionist historians maintain that the aged in nineteenth-century America and north-western Europe usually preferred to reside alone or with only their spouse. According to this interpretation, the aged ordinarily resided with their adult children only out of necessity, especially in eases of poverty or infirmity. This article challenges that position, arguing that in mid-nineteenth-century America coresidence of the aged with their children was almost universal, and that the poor and sick aged were the group most likely to live alone. The article suggests that the decline of the multigenerational family in the twentieth century is connected to the rise of wage labour and the diminishing importance of agricultural and occupational inheritance.

Incomplete information is frequent in real-world applications. This is often the case in data integration scenarios, in scientific data collections, or whenever the information is acquired using human interaction and is erroneous or imperfect. The different interpretations of incomplete information yield different possible worlds. A system for managing incomplete data faces the challenge of being able to represent large sets of possible worlds compactly, while at the same time supporting efficient processing of the data. Nevertheless, there has been little research so far into expressive yet scalable systems for managing incomplete information. Most current representation models have at least one of two flaws; some of them (such as or-set relations and v-tables) are not strong enough to represent query answers within the same formalism...

Concern is often voiced about the quality ofAmerican schoolteachers. This paper suggests that, while the relative quality of teachers is declining, this decline may be the result of technological changes that have raised the priceof skilled workers outside teaching without affecting the productivity of skilled teachers. Growth in the price of skilled workers can cause schools to lower the relative quality of teachers and raise teacher quantity instead. Evidence from the NLSY demonstrates that wage and schooling are good measures of teacher quality. Analysis of US Census microdata then reveals that the relative schooling and experience-adjusted relative wagesof US schoolteachers have fallen significantly from 1940 to 1990. Moreover, class sizes have also fallen substantially. The declines in class size and in relative quality seem correlated over time and space with growth in the relative price of skilled workers.

Legislation in the 1970s created a Notch in social security payments, with those born after January 1, 1917, receiving sharply lower benefits. Using restricted-use versions of the National Mortality Detail File combined with Census data, we use this quasi experiment to examine the income mortality link in an elderly population. Estimates from difference-in-difference and regression discontinuity models show the higher-income group has a statistically significantly higher mortality rate, contradicting the previous literature. We also found that younger cohorts responded to lower incomes by increasing postretirement work effort, suggesting that moderate employment has beneficial health effects for the elderly.

Although older adults comprise the most rapidly growing segment of the American population, research on aging in Appalachia is surprisingly limited. For this reason, we take this opportunity to consider aging and the life course in an Appalachian context beginning from what is known, and elaborating on it as a way of identifying potentially fruitful knowledge gaps and future directions. Often, average figures mask more complex ways in which the Appalachian and non-Appalachian regions differ, many of which have implications for our understanding of aging and the life course.

Women are currently the majority of U.S. college students and of those receiving a bachelors degree, but were 39 percent of undergraduates in 1960. We use three longitudinal data sets of high school graduates in 1957, 1972, and 1992 to understand the narrowing of the gender gap in college and its reversal. From 1972 to 1992 high school girls narrowed the gap with boys in math and science course taking and in achievement test scores. These variables, which we term the proximate determinants, can account for 30 to 60 percent of the relative increase in womens college completion rate. Behind these changes were several others: the future work expectations of young women increased greatly between 1968 and 1979 and the age at first marriage for college graduate women rose by 2.5 years in the 1970s, allowing them to be more serious students. The reversal of the college gender gap, rather than just its elimination, was due in part to the persistence of behavioral and developmental differences between males and females.

This paper proposes and estimates a static transferable utility model of the marriage market. The model generates a nonparametric marriage matching function with spillover effects. It rationalizes the standard interpretation of marriage rate regressions and points out its limitations. The model was used to estimate U.S. marital behavior in 1971/72 and 1981/82. The marriage matching function estimates show that the gains to marriage for young adults fell substantially over the decade. Unlike contradictory marriage rate regression results, the marriage matching function estimates showed that the legalization of abortion had a significant quantitative impact on the fall in the gains to marriage for young men and women.

A number of previous studies found interesting links between the early-life experiences and a subsequent mortality in later life. Such findings justify further advancement of these studies by exploring possible links between childhood conditions and exceptional longevity (survival to 100 years). However a number of methodological issues have to be resolved (such as data availability, quality, as well as research approaches), before starting a comprehensive research project on childhood predictors of exceptional longevity. This paper represents an attempt to do a preliminary study of related methodological issues in order to ensure feasibility of the subsequent large-scale research efforts. The following questions are explored in this study: Where to get data on exceptional longevity and childhood conditions? What is the quality of this data, and how can this data quality be checked and improved? What methods of data analysis to use? What are the hypotheses to test, and what are the preliminary findings to validate?The study presents a detailed analysis of available resources on exceptional longevity and childhood conditions, a rigorous evaluation of data quality and testing different approaches to improve the quality of the data. As a result of these methodological explorations the following multi-step procedure of data collection and data cleaning has being suggested and tested in practice:Step 1. To extract data on alleged centenarians and their childhood conditions from computerized genealogies, pre-selected on the basis of their expected sufficient data quality (a number of indicators of data quality has been identified for this purpose). Following this procedure we extracted detailed family data for 991 alleged centenarians born in 1875-1899 in the United States from publicly available computerized genealogies of 75 million individuals identified in our previous study (Gavrilova, Gavrilov,1999).Step 2. To validate the claims of exceptional longevity by cross-checking these records with the US Social Security Administration database on deceased person for the entire US population.Step 3. To double-check the validated centenarian claims again for the accuracy of birth date information by matching the records with the early US Censuses (1900, 1910, 1920 and 1930).Step 4. To enhance further the quality of records with confirmed exceptional longevity by adding information on childhood conditions available in the early US Censuses. This multi-step procedure of data collection and quality evaluation is described in this paper in great detail, to demonstrate how this approach works in practice. A new validated dataset on exceptional longevity and childhood conditions has been successfully developed in this study using this multi-stepprocedure of data collection and validation. This validated dataset contains information on 485 centenarians born in the United States in 1890-1900 and the families where they were raised. Thus, this multi-step approach has been tested in practice and it could be recommended for a subsequent large-scaleresearch project on childhood predictors of exceptional longevity. At the next step of this study we used the collected and validated dataset on exceptional longevity and childhood conditions as a sandbox for applying different methods of data analysis and testing a number of specific hypotheses about childhood predictors of exceptional longevity. Specifically, we followed theearlier studies by Preston and Haines (1991) who found that the lowest sickness burden in early life(measured through the level of child mortality) was observed in the families of farmers, and families living in the Western region of the United States in the 1900s. This leads to a testable prediction that centenarians should occur more often in the families of farmers, and in families living in the Western region of the United States, if more healthy childhood conditions are indeed conductive for exceptionallongevity later in life. We tested these hypotheses using a method of multiple logistic regression to compare household characteristics of 'centenarian' families in the 1990s with household characterstics of representative sample of the 1900 US Census (IPUMPS data). Indeed it turned out that centenarians were born more often in the families of farmers, and in families living in the Western region as predicted by the 'childhood conditions' hypothesis.We also explored different approaches to study the effects of the birth order on exceptional longevity, starting with analysis of a simple summary statistic like the 'centenarian birth order ration' and the 'centenarian birth order difference', followed by more sophisticated analysis based on multiple logistic regression. These studies revealed that there is a statistically significant association between the chances of exceptional survival and the birth order.Finally, a method of within-family analysis has been applied to investigate the occurrence patterns for centenarians among siblings, which allows researchers to avoid confounding caused by between-family variation. This approach was implemented using conditional logistic regression with a binary outcome variable describing either a centenarian, or non-centenarian survival outcome. For this in-depth analysis the 198 validated centenarians born in USA in 189-1893 were identified, and their family histories were reconstructed using the US Censuses, the US Social Security Administration database, genealogical records and other supplementary data resources. The following predictor variables were explored: sex, birth order, paternal age at person's birth, maternal age at person's birth, and the season of birth.The study found that first-born siblings are more likely to become centenarians when compared to later-born siblings (odds ration=1.77, 95% CI=1.18-2.66, P=0.006). This protective effect of first-born status can not be simply explained by differences in child mortality, because it persists when a comparison is made with those siblings only who survived to adulthood (age 20): odds ration=1.95, 95% CI=1.26-3.01, P=0.003. Moreover, even at age 75 it still helped to be a first-born child in order to become a centenarian (odds ration=1.66, 95% CI=1.02=2.69, P=0.04).In order to find out the mechanism of the birth-order effect, a multivariate analysis with included parental variables was performed. This multivariate analysis found that the protective effect of being first-born is driven mostly by the young maternal age at person's birth 9being born to mother younger than 25 years). Being born to young mother is the major predictor of human longevity (odds ratio=2.03, 95% CI=1.33-3.11, P=0.001). Moreover, even at age 75 it is still important to be born to young mother in order to survive to 100 years (odds ratio=1.87, 95%CI=1.15-3.05, P=0.01).The results of this study demonstrate that childhood conditions are indeed very important in determining the chances of exceptional longevity and justify the feasibility of the subsequent large-scale research efforts in this direction.

This paper investigates how immigrants in the United States respond to changes in dual citizenship laws in their origin country. In the 1990s Colombia, the Dominican Republic, Ecuador, Costa Rica and Brazil revoked the previous rule that took away nationality of the origin country from those who became citizens of another country. Using data from the 1990 and 2000 censuses, I find a sizable and statistically significant effect of granting dual citizenship on the probability of naturalization in regressions that include controls for other factors (such as welfare reform) that changed the incentives to naturalize over the 1990s. Immigrants recently granted dual nationality rights also experience employment gains, but no earnings gains. The effects of dual citizenship on labor outcomes, when interpreted through naturalization, are consistent with American citizenship providing greater employment opportunities, and a more rapid wage growth that might not have shown its effects yet among recently naturalized immigrants.

What are the consequences of resource-based regional specialization, when it persists over a long period of time? While much of the literature argues that specialization is beneficial, recent work suggests it may be costly in the long run, due to economic or political reasons. I examine this question empirically, using exogenous geological variation in the location of subsurface oil in the Southern United States. I find that oil abundant counties are highly specialized: for many decades their mining sector was almost as large as their entire manufacturing sector. During the 1940s and 1950s, oil abundant counties enjoyed per capita income that was 20-30 percent higher than other nearby counties, and their workforce was better educated. But whereas in 1940 oil production crowded out agriculture, over the next 50 years it caused the oil abundant counties to develop a smaller manufacturing sector. This led to slower accumulation of human capital in the oil abundant counties, and to a narrowing of per capita income differentials to about 5 percentage points. Despite this caveat, the gains from specialization were large, and specialization had little impact on the fraction of total income spent by local government or on income inequality.

This paper studies the nonparametric identification and estimation of a generalized Roy model that includes a non-pecuniary component of utility associated with each alternative. This generalized model allows amenity or risk considerations to affect occupation choice in the classic Roy model. More generally, such a generalization would be useful in any economic setting where individuals place value on more than a single outcome of interest when choosing among discrete treatments or behaviors. The starting point for our study of identification is the well-known result for the pure Roy model in Heckman and Honore (1990): that any crosssectional dataset can be rationalized by underlying population wage distributions in which wage offers are distributed independently across sectors -- i.e., that the correlation of wage offers across sectors is unidentified. A common interpretation of this important result is that, without parametric restrictions or the availability of covariates, all of the useful content of a cross sectional dataset is absorbed in a restrictive specification that imposes independence. While this is certainly true within the pure Roy model, we demonstrate that it is in fact possible to identify, under relatively innocuous support assumptions, a common non-pecuniary component of utility associated with each sector. These results apply even in empirical settings characterized by many alternative sectors and without the need for additional covariates. We develop nonparametric estimators corresponding to two alternative assumptions under which we prove identification, describe their asymptotic properties and provide Monte Carlo evidence on their performance in small samples. We close the paper by applying our preferred estimator to study migration across US labor markets using data drawn from the 2000 US Census.