For over thirty years, research on Spanish in the U.S. has demonstrated an inexorable loss of the language among Spanish speaking populations. This study shows, however, that analyses of 1980, 1990, and 2000 U.S. Census data, using an innovative approach known as a synthetic cohort analysis, reveal a high degree of transmission of Spanish from first generation to second generation speakers. For the purpose of tracking reported language use of individuals starting at ages 57 and ending at ages 1517, data from the Integrated Public Microdata Series are used here to create two simulated longitudinal samples of Spanish speakers over a ten-year period. English language acquisition is also examined, and the results indicate that second generation speakers are bilingual, with a high degree of control of both Spanish and English.

We construct and test a new model of employer-provided health insurance provision in the presence of adverse selection in the health insurance market. In our model, employers cannot observe the health of their employees, but can decide whether to offer insurance. Employees sort themselves among employers who do and do not offer insurance on the basis of their current health status and the probability distribution over future health status changes. We show that there exists a pooling equilibrium in which both sick and healthy employees are covered as long as the costs of job switching are higher than the persistence of health status. We test and verify some of the key implications of our model using data from the Current Population Survey, linked to information provided by the U.S. Department of Labor about the job-specific human capital requirements of jobs.

Increased immigration coupled with the relative youth of the foreign born is fueling rapid growth in the number of immigrant children, i.e., foreign born or first generation and U.S.-born children of foreign-born parents, the second generation, attending U.S. schools. Policymakers need to understand how school persistence varies by generation status and which individual characteristics matter for high school completion. California census data suggest significant heterogeneity in enrollment propensities by gender (Garvey, 2005). Women are on average about 20 percent more likely to be in school than young men, with the strongest advantage accruing to the second generation. I exploit the large sample sizes of the 2000 IPUMS data to analyze gender differences in school enrollment and labor force participation of first and second generation youth relative to natives whose parents were also born in the United States (the third+ generation), as well as by disaggregated race/ethnicity/national origin groups. A key contribution of this paper is its improvement of IPUMS-constructed variables that link children to resident parents and thus define a childs family context. My algorithms use household structure information to determine a youths generation status.The census reports completed schooling, school participation status, and labor force participation. I use logistic regression models to analyze the interplay of gender and generation status (controlling for individual, family, and neighborhood characteristics) on school participation for youth ages 16 to 18. Multinomial logistic regression is then used to explicitly examine the pathways out of schooling by analyzing how the relative attractiveness of schooling versus competing activities of labor market work and idleness vary by gender and generation status. I then examine how the impact of gender varies by race/ethnicity across generation status groups for these two sets of outcomes.

Widows and Orphans First" investigates the importance of local economies and values in the origins of the welfare state through an exploration of widows' lives in three industrial American cities with widely differing economic, ethnic, and racial bases. In Fall River, Massachusetts, employment was regarded as the solution to widows' poverty, so public charitable expenditure was drastically limited.In Pittsburgh, where few jobs were available for women or children - and where jobs for men were in 'widowmaking' industries such as steel and railroading - the city's charitable establishments were more sympathetic. In the border city of Baltimore, which had a large African American population and a diverse economy that relied on inexpensive child and female labour, funds for public services were limited, and African Americans tended to establish their own charitable institutions. In this unique comparative study of widows' welfare and family economy, Jay Kleinberg examines the role of children in society and the development of social welfare policy for widows.

Introduction: This study examines demographic and socioeconomic aspects of the Mexican population of the New York City area from 1990-2005. Methods: Data on Latinos and other racial/ethnic groups were obtained from the U.S. Census Bureau American Community Survey, reorganized for public use by the Minnesota Population Center, University of Minnesota, IPUMSusa. Cases in the dataset were weighted and analyzed to produce population estimates. Results: The Mexican-origin population of New York City was the city’s fastest-growing Latino national group between 1990 and 2005. From a population of 55,587 in 1990 Mexicans increased to 183,792 in 2000 and 227,842 in 2005.1 By 2005 Mexicans had become the third largest Latino nationality in NYC behind Puerto Ricans (790,609) and Dominicans (570,641). The yearly growth rate between 1990 and 2000 among Mexicans was 12.7% although this slowed to 4.4% between 2000 and 2005.The Dominican population of NYC increased at 1.4% yearly between 2000 and 2005 and the Puerto Rican population experienced their first decline ever, falling slightly by - 0.2% annually over the same period. Mexicans were a highly stratified Latino national group and in this sense were no different from other Latino nationalities. About 20% of all households earned less than $20,000 annually over the 15 year period measured in this report. At the top of the social hierarchy there was an increase in the percentage of Mexican households earning more than $75,000 annually and this is indicative of opportunities for upward social mobility within NYC for better educated and better skilled Mexicans. Discussion: Work force data suggest two significant differences between Mexicans and other Latino nationalities in the City. The first is an extraordinarily higher percentage of men who were employed in 2005 and a correspondingly low percentage of men who were employed or out of the work force. The second is the comparative low rates of low both unemployed working-age Mexican females who were working in 2005 compared with women in other Latino national groups. As indicated above, this may have been due to child-rearing responsibilities connected to the extraordinarily high birth rates found among Mexican females in comparative perspective.

Immigrants arrive in the U.S. with better than average health, which declines over time. Clinical preventive services can prevent or delay some of that decline, but little research in this area focuses specifically on Mexican immigrants who are the largest contemporary immigrant group. This article finds that recent Mexican immigrants were the least likely to receive preventive care services, even after adjusting for sociodemographic differences in the population. Long-stay Mexican immigrants were more similar to U.S.-born Mexican Americans in preventive service use rates, who in turn had lower rates than U.S.-born non-Latino whites. Monolingual Spanish speaking Mexican immigrants were the least likely to have obtained preventive services. Having no usual source of care is the strongest predictor of the underuse. The persistent gap in preventive services across all subgroups of adults of Mexican origin suggests structural barriers to their preventive care.Keywords Emigration and immigration - Hispanic Americans - Health promotion - Health services accessibility - Preventive Services

This paper examines the extent to which Mexican emigrants to the United States are negatively selected, that is, have lower skills than individuals who remain in Mexico. Previous studies have been limited by the lack of nationally representative longitudinal data. This one uses a newly available household survey, which identifies emigrants before they leave and allows a direct comparison to non-migrants. I find that, on average, US bound Mexican emigrants from 2000 to 2004 earn a lower wage and have less schooling years than individuals who remain in Mexico, evidence of negative selection. This supports the original hypothesis of Borjas (AER, 1987) and argues against recent findings, notably those of Chiquiar and Hanson (JPE, 2005). The discrepancy with the latter is primarily due to an under-count of unskilled migrants in US sources and secondarily to the omission of unobservables in their methodology.

Some anti-discrimination laws have the perverse effect of harming the very class they were meant to protect. This paper provides evidence that age discrimination laws belong to this perverse class. Prior to the enforcement of the federal law, state laws had little effect on older workers, suggesting that firms either knew little about these laws or did not see them as a threat. After the enforcement of the federal Age Discrimination in Employment Act (ADEA) in 1979, white male workers over the age of 50 in states with age discrimination laws worked between 1 and 1.5 fewer weeks per year than workers in states without laws. These men are also .3 percentage points more likely to be retired and .2 percentage points less likely to be hired. These findings suggest that in an anti-age discrimination environment, firms seek to avoid litigation through means not intended by the legislation -- by not employing older workers in the first place.

Nonparametric methods have become increasingly popular in statistics and probabilistic AI communities. One well-known nonparametric method is nearest neighbor. It uses the observations in the training set T closest in input space to a query q to form the prediction of q. Specifically, when k of the observations in T are considered, it is called k-nearest-neighbor (or k-NN). Despite its simplicity, k-NN and its variants have been successful in many machine learning problems, including pattern recognition, text categorization, information retrieval, computational statistics, database and data mining. It is also used for estimating sample distributions and Bayes error.However, k-NN and many related nonparametric methods remain hampered by their computational complexity. Many spatial methods, such as metric trees, have been proposed to alleviate the computational cost, but the effectiveness of these methods decreases as the number of dimensions of feature vectors increases. From another direction, researchers are trying to develop ways to find approximate answers. The premise of this research is that in many cases it is not necessary to insist on the exact answers; instead, determining an approximate answer should be sufficient. In fact, some approximate methods show good performance in a number of applications, and some methods enjoy very good theoretical soundness. However, when facing hundreds or thousands dimensions, many algorithms do not work well in reality.I propose four new spatial methods for fast k-NN and its variants, namely KNS2, KNS3, IOC and spill-tree. The first three algorithms are designed to speed up k-NN classification problems, and they all share the same insight that finding the majority class among the k-NN of q need not to explicitly find those k-nearest-neighbors. Spill-tree is designed for approximate k-NN search. By adapting metric-trees to a more flexible data structure, spill-tree is able to adapt to the distribution of data and it scales well even for huge high dimensional data sets. Significant efficiency improvement has been observed comparing to LSH (localify sensitive hashing), the state of art approximate k-NN algorithm. We applied spill-tree to three real-world applications: shot video segmentation, drug activity detection and image clustering, which I will explain in the thesis.

The theory of ability misperception posits that employers will offer greater rewards to whites than nonwhites for similar levels of prior experience (Proposition 1) but that racial/ethnic differences in the return to additional tenure or seniority with the current employer will be smaller (Proposition 2). To advance the existing empirical literature, this paper evaluates the validity of these propositions by using data on black, Latino, and white workers drawn from the Multi-City Study of Urban Inequality. The analysis is conducted separately for women and men and controls for a wider range of workplace setting descriptors than was used in previous studies. Our results offer support for Proposition 1 and for Proposition 2. We find that nonwhites, regardless of job setting, receive relatively poor returns to prior workplace experience (the lone exception is Latinas). Second, nonwhites typically receive greater wage gains for accumulating additional tenure than whites.

Some anti-discrimination laws have the perverse effect of harming the very class they were meant to protect. This paper provides evidence that age discrimination laws belong to this perverse class. It exploits an unusual aspect of the policy for enforcement of the federal 1968 Age Discrimination in Employment Act (ADEA), which made filing an age discrimination claim less burdensome in some states than in others. After the enforcement of the federal law, white male workers over age 50 in states where the federal government allowed 300 days to file a discrimination complaint worked between 1 and 1.5 fewer weeks per year than did workers in states without laws. These men were also .3 percentage points more likely to be retired and .2 percentage points less likely to be hired. These findings suggest that in an anti-age discrimination environment, firms seek to avoid litigation through means not intended by the legislationby not employing older workers in the first place.

This paper is about non-approximate acceleration of high-dimensional nonparametric operations such as k nearest neighbor classifiers. We attempt to exploit the fact that even if we want exact answers to nonparametric queries, we usually do not need to explicitly find the data points close to the query, but merely need to answer questions about the properties of that set of data points. This offers a small amount of computational leeway, and we investigate how much that leeway can be exploited. This is applicable to many algorithms in nonparametric statistics, memory-based learningand kernel-based learning. But for clarity, this paper concentrates on pure k-NN classification. We introduce new ball-tree algorithms that on real-world data sets give accelerations from 2-fold to 100-fold compared against highly optimized traditional ball-tree-based k-NN . These results include data sets with up to 106 dimensions and 105 records, and demonstrate non-trivial speed-ups while giving exact answers.

The impact of agglomeration and spatial concentration uponeconomic productivity has been the focus of many studies.Using estimatedwage equations and geographic data, this research focuses on the reduction ofagglomeration economies and human capital externalities. The analysis of the data allows researchers to evaluate the impact ofagglomeration distance upon labor productivity.While agglomeration isexamined using concentric rings from the workplace, productivity is measuredaccording to existing wages.Geological characteristics are alsoconsidered significant for understanding measurement error. Influential factors, such as education level, human capital, and the urbanwage premium, are considered.Previous literature regarding agglomeration,productivity, and wages is explored.The data set was primarily drawn fromthe 5% Integrated Public Use Microdata Series (IPUMS) from 2000, as well as2000 U.S. Census data. The findings of the paper indicate that evidence supporting an urban wagepremium exists. The analyses further suggest that human capital positivelyimpacts agglomeration. Further, a close proximity to college-educated workers is correlated with anincrease in productivity and wages; the proximity to less-educated workers iscorrelated with an opposite effect.The effect of human capital spilloversindicate a positive influence, extending as far as 50 to 100 miles. (AKP)