PRODUCTIVITY OF HIGHLY SKILLED IMMIGRANTS: ECONOMISTS IN THE POSTWAR PERIOD.

MCDOWELL, JOHN M. ; SINGELL, JR, LARRY D.

LARRY D. SINGELL, JR [*]

Prior work finds declining immigrant quality in the postwar period that is linked to source-country and skill-composition changes associated with the 1965 Immigration Act. This paper uses a unique panel of foreign- and native-born American Economic Association members to show that the highly skilled experienced a similar shift away from European migrants toward those from Asia. However, the findings do not indicate that this change in source-country composition has been accompanied by a decline in quality; rather, the most recent cohorts of foreign-born economists appear to be more productive than their native counterparts. (JEL J61)

I. INTRODUCTION

This article is the first systematic analysis of the attributes and abilities of foreign-born U.S. economists since a descriptive study of the brain drain by Grubel and Scott [1967]. It also provides some of the first formal evidence for the largely theoretical proposition in earlier work by Bhagwati and Rodriquez [1975] that the United States draws the best skilled workers from abroad. However, recent immigration literature has not focused on the brain drain but has generated considerable debate concerning whether the 1965 Immigration and Nationality Act has reduced the skill quality of U.S. immigrants. In particular, descriptive evidence in Borjas [1987] and Greenwood [1983], respectively, indicates that the 1965 law shifted the composition of U.S. immigrants toward non-European nations and increased the number of low-skilled migrants to the United States. Our article extends this work by examining whether there has been a post-1965 change in the source-country mix of Ph.D. economists and whether such chan ges impact the productivity of the foreign born in comparison to their native counterparts.

Most empirical studies of immigrant productivity have focused on earnings. Early cross-sectional studies of immigrant earnings, including Chiswick [1978a], concluded that the age-earnings profile of immigrants is steeper than that of comparable native-born workers and, though initially lower, crosses the native profile 10 to 15 years after immigration. These results are consistent with the widely held belief that the immigration process self-selects the highly motivated and industrious from their respective source countries. Thus immigrants, although they initially enter the United States with relatively low levels of human capital, eventually become more productive than their native counterparts.

More recent work by Borjas [1985, 1987] has contested these conclusions, arguing that it is not possible to separately identify assimilation and cohort effects using a single cross-sectional data set. In particular, the positive correlation between earnings and years since migration in cross-sectional data may be due to immigrant assimilation or a cohort effect that arises from the declining skill quality of more recent immigrant groups. Borjas combines multiple cross-sectional data sets to separately identify assimilation and cohort effects and finds evidence indicating that earlier cohorts of immigrants do relatively better than both natives and more recent immigrant cohorts, suggesting a decline in immigrant quality. Although Chiswick [1986] challenges the contention of declining immigrant quality and Funkhouser and Trejo [1995] find evidence that immigrant quality may have increased during the 1980s, Borjas [1995] contends that the empirical evidence consistently shows that immigrant quality has declined for most of the postwar period.

However, the decline in immigrant quality has not necessarily been found within specific groups of immigrants. In particular, Greenwood and McDowell [1986] find that the change in skill quality is related to the shift in the source-country composition of immigrants away from European nations toward those from Asia and Latin American after the 1965 Immigration Act. Thus, explanations for the relatively poor performance of recent immigrants, including those in LaLonde and Topel [1991, 1992) and Borjas [1992], focus on the post-1965 changes in the incentives to migrate to the United States that have increased the number of low-skilled, non- European immigrants.

This article takes an alternative approach by focusing on a group of highly skilled workers (i.e., Ph.D. economists). In particular, we use publication and personal data for a panel of native- and foreign-born economists who are members of the American Economics Association (AEA). Our analysis finds a source-country composition shift for the economics profession over the last 60 years that is similar to the broader population of immigrants. We use a tobit model of research productivity to examine whether, consistent with prior analysis in Borjas [1995], the shift in the national origin mix away from traditional European source countries has generated a less "successful" immigrant flow.

In the article, traditional cross-sectional specifications that focus on assimilation effects are compared with longitudinal specifications that distinguish between assimilation and cohort effects. In addition, our unique data permit us to extend prior work by using both detailed personal/regional controls and panel techniques to examine whether observed assimilation and cohort effects can be attributed to changes in either observed and/or unobserved heterogeneity, respectively. This permits us to examine more clearly whether there are changes in the relative productivity of foreign-born economists over time. Our results provide evidence of systematic productivity differences between highly skilled foreign- and native-born workers that varies over time in response to both immigration policy and market conditions.

II. THE PANEL

The analysis uses a panel of foreign- and native-born American Economic Association (AEA) members who entered the economics profession over a period prior to World War II and ending in the mid-1980s. These unique data, though contained in readily available sources, have not previously been organized in a way that could be empirically exploited. Thus, it is useful to examine briefly how the panel is constructed to study changes in the characteristics of foreign-born AEA members, both in absolute terms and in relative comparison to the native born, over a time period that encompasses the passage of the 1965 Immigration Act.

Construction of the Panel

The primary sources for the panel is the biographical listings in the 1964 and 1974 AEA directories and the 1985 computer tape, which are used to identify the native- or foreign-born status of AEA members. [1] We select all foreign-born AEA members aged 65 or less if the individual received a Ph.D. (or equivalent degree) from: (1) a U.S. institution, or (2) a foreign institution, but works for an employer located in the United States. [2] The native-born control group must also be aged 65 or less and meet criterion (1) or (2). The native sample group is initially selected at a 50% rate of the foreign-born population within each of the respective 1964, 1974, and 1985 cross-sections. For all individuals, demographic and career-related data are collected from the biographical listings in various AEA directories. [3] These data include date and place of birth, gender, degree years and source institutions, principal employment and (if foreign born) the year of entry into U.S. employment, and the individual's majo r fields of interest. [4]

The panel is constructed in several steps. First, the foreign-born panel is obtained by simply cross-checking and linking the names of individuals who match in each of the cross-sections. This yields a panel that traces the careers of 65% of foreign-born economists in more than one cross-section. Next, the initial native-born control group is supplemented by adding observations for the originally sampled native-born economists in each of the other cross-sections in which they are also found. For example, the name of a native-born economist in the initial 1974 random sample is cross-checked and observations added in other cross-sections if this individual also is observed in the 1964 or 1985 AEA directories. Because the overlap of the initial random cross-sectional samples of native-born economists for 1964, 1974, and 1985 is relatively small, the number of native-born economists in the final panel data actually exceeds that of the foreign born even though the native born are originally selected at a 50% rate of the foreign born in each cross-section. Nonetheless, following all cross-sectional observations of the control group over time is consistent with the construction of the foreign-born panel and ensures a sizable panel that includes approximately 65% of the native-born economists originally sampled.

The final step in constructing the panel involves adding data relating to a 1989 cross-section. The 1989 cross-section is constructed by taking all foreign- and native-born individuals in their respective 1985 cross-section and including individuals in the 1989 data if they otherwise met the criteria for sample inclusion. Thus, no attempt is made to add "new" names in the 1989 cross-section. [5] By adding the 1989 cross-section, the panel data include individuals who received their Ph.D. after 1974 and include the most recent year for which all the data necessary for the analysis of research output are available.

The personal data are supplemented with information pertaining to publication activity that is collected from various issues of the ABA Index of Economic Articles. Publications are recorded over a 2-year period and then aggregated to provide a measure of research output. More specifically, an individual's publications are represented by articles that are published in the year of the cross-sectional observation plus articles published in the year immediately following (i.e., 1964-65, 1974-75, 1985-86, or 1989-90). [6] The publication data also include information on the number of coauthors and journal placement. This latter information is used to provide alternative weighted measures of publication activity. [7]

The use of panel data follows the precedent established in Huang [1987] and Boijas [19851 that has shifted the focus from a single cross-section of data to longitudinal data constructed by aggregating cross-sectional random samples of different sets of individual in several time periods. However, our panel offers a number of advantages over prior analyses. First, while recent analyses have used longitudinal data, most do not follow the same individuals over time, as permitted by our AEA data. [8] Without a panel, longitudinal studies are not able to separate aggregate changes in cohort ability (e.g., less able persons migrate to the United States) from changes in the unobserved quality of time varying immigrant attributes (e.g., a decline in the quality of a foreign versus a U.S. education). This is particularly a problem because the source country for the time-varying attributes, such as education, is generally unknown. Second, prior studies use earnings that are only an indirect measure of productivity. Ho wever, our publication measure is a direct qualitative measure of output. Finally, the homogeneity of economists, although limiting generalizations to broader populations, permits us to examine cohort effects for workers who have similar training and opportunities.

Descriptive Statistics

Descriptive statistics for foreign- and native-born AEA members who enter the profession before and after 1965 are included in Table I. The data include 2,958 person-year observations of 1,180 foreign-born economists and 3,385 person-year observations of 1,354 native-born economists. Thus, on average, both native- and foreign-born economists contribute roughly 2.5 panel years to the data. Overall, the descriptive statistics suggest that foreign-born economists are older with more experience, and they are more likely to be male and placed in academia than their native-born counter-parts. Foreign-born economists also are less likely to attend a top-35 Ph.D. institution, but this may simply reflect that their degrees are more frequently conferred by institutions outside the United States.

Table I suggests a number of differences in the profession before versus after 1965. The most notable change is the source-region mix of the foreign born. In particular, among those who entered the economics profession prior to 1965, over 70% of foreign-born economists migrated from Europe or Canada, whereas these regions contributed only 40% of foreign-born economists entering after 1965. At the same time, foreign-born economists from Asia nearly doubled, from 20% to 37%, between the two periods. The most distinct difference between the post-1965 source-region mix of economists and that of the general population of U.S. immigrants is the relatively small representation of economists from Latin America after 1965. [9] Our empirical analysis examines if there are changes in relative productivity over time after controlling for observed and unobserved heterogeneity of native- and foreign-born economists.

III THE EMPIRICAL MODEL

The empirical model encompasses the traditional cross-sectional approach, subsequent modifications by longitudinal studies, and more recent attempts to explain observed cohort effects by including more detailed controls for worker and country attributes. In addition, we extend prior analyses by the introduction of panel techniques to control for unobserved worker heterogeneity. A general tobit model is developed to show how the specifications are related.

Tobit Model of Research Productivity

Economists, whether foreign or native born, are viewed as allocating time and resources to maximize utility over their career. The utility of professional economists is likely to depend on research output both because it enters as an argument in their utility function and because the institutions for which they work compensate them to various degrees for research activities. Thus, utility maximization is postulated to generate an optimal amount research for an economist i at time interval t, [[Q.sup.*].sub.it] that can be expressed as a linear function:

(1) [[Q.sup.*].sub.it] = [alpha][X.sub.it] + [beta][Y.sub.it] + [gamma][Z.sub.it] + [[epsilon].sub.it]

where [X.sub.it] is the vector of variables included in traditional cross-sectional studies; [Y.sub.it] is a vector of cohort effects found in longitudinal studies; [Z.sub.it] is a vector observable ability measures that include regional dummies and personal characteristics; [alpha], [beta], and [gamma] are parameter vectors; and [[epsilon].sub.it] is the error term.

Like many optimization problems, the optimal amount of research consumed (produced) is not restricted to be greater than zero (e.g., the labor-leisure model). Nonetheless, observed research output, [Q.sub.it], must be nonnegative. Using equation (1), [Q.sub.it] can be described by a semidiscrete function:

(2.1) [Q.sub.it] = 0 [sim] if [[epsilon].sub.it] [less than or equal to] ([alpha][X.sub.it] + [beta][Y.sub.it] + [gamma][Z.sub.it])

(2.2) [Q.sub.it] = [[Q.sup.*].sub.it] [sim] if [[epsilon].sub.it] [greater than] ([alpha][X.sub.it] + [beta][Y.sub.it] + [gamma][Z.sub.it]).

Expressions (2.1) and (2.2) form the basis for several specifications of a tobit model of the research output, which differ from one another based on the assumptions regarding the parameters and/or the error term.

The Empirical Specification

As a point of departure we use the number of published pages to measure [Q.sub.it] because it has relatively more variation than its alternatives. Nonetheless, the other publication measures are also used and show that the results are robust to the choice of dependent variable. Our base specification follows the early cross-sectional work that assume [beta] and [gamma] equal zero and includes measures of human capital (i.e., age and experience) and, for immigrants, the differential return to post-Ph.D. U.S. experience. For all economists, research output is expected to increase with post-Ph.D. experience, whereas its square and cubic are included to account for possible nonlinearities in research output documented in Oster and Hamermesh [1998] and McDowell [1982]. The individual's age and its square are also included to account for the depreciation of human capital that occurs, if nothing else, because technical skills eventually become dated. The number of years since entry, as measured by years since first post-Ph.D. U.S. employment, and its square are used to proxy for the assimilation effect. The inclusion of the years-since-entry variable roughly follows the methodology found in Borjas [1990] and Chiswick [1978a]. Unfortunately, the AEA directories include information on an economist's first U.S. appointment, but not when they actually first entered the United States. Thus, the years-since-entry variable measures the immigrant-specific return to post-Ph.D. U.S. experience.

Contrary to early cross-sectional work, our data include multiple time periods and male and female workers. Thus, [X.sub.it] includes two additional variables. First, to account for possible gender differences in research output documented in Hansen et al. [1978], the model includes a binary variable that equals one for female economists. Second, following longitudinal studies that use multiple cross-sections, a dummy variable is included for the 1974, 1985, and 1989 cross-sections to test for possible period effects relative to the 1964 sample.

To make use of the longitudinal aspects of our data, several specifications supplement [X.sub.it] with [Y.sub.it] that includes 10 entry cohort dummies for the 5-year intervals, starting with the first observed cohort in 1936-40 and ending with the last observed cohort in 1981-85. The 5-year interval ensures at least 30 observations in each interval and has the added advantage of being directly comparable to the time intervals used in prior longitudinal studies. These entry cohort variables permit us to observe whether the quality of foreign-born economists has changed over time.

Prior work suggests that the observed cohort effects for a random sample of immigrants can be attributed to changes in the source country mix of the foreign born and their associated differences in observed attributes. This hypothesis is tested by introducing the vector [Z.sub.it] that includes two groups of variables that broadly proxy for immigrant quality. First, a set of 10 binary variables for region of origin are used and include (in alphabetical order) Africa, Asia, the Caribbean, Eastern Europe, Mexico and Central America, the Middle East, Northwest Europe, Oceania (i.e., Australia, New Zealand, and the Pacific Islands), South America, and Southwest Europe. Canada is used as the excluded region because it includes a large number of immigrants across all time periods and because of its cultural similarity and spacial proximity to the United States. Although the actual country of origin is known for each of the foreign-born economists, the relatively broad regions ensure a reasonable number of observati ons in each category and conserve degrees of freedom.

Second, research output may differ over time and by country because of changes in: (1) the occupational distribution toward jobs that are less research oriented (i.e., from academia to government), (2) the quality of education with relatively fewer graduates from top Ph.D. programs, or (3) the choice of field toward those that produce less research or those that have experienced less growth in research opportunities. Thus [Z.sub.it] also includes 3 measures for occupation type, 4 schooling measures, and 15 field-of-specialization variables that are used to examine whether these observed differences can account for possible cohort or regional differences among immigrants. [10]

The model in its most general form permits the productivity of the foreign born to vary over time via experience (including a differential return to U.S. experience), period, and cohort effects. However, because there are only two sources of variation in the model (i.e., time and individual), all of these effects cannot be separately identified for the foreign born. Following the identification strategy in Borjas [1985], the native- and foreign-born samples are combined and the model is estimated with the restriction that the human-capital measures (except for years since entry), the personal attributes (except for the entry-cohort variables), and period effects are the same for native-born and foreign-born economists. The period effects capture overall changes in the tendency to publish in the profession over time, which may reasonably be assumed to be the same for native-born and foreign-born economists. [11]

The Error Term

The error term, [[epsilon].sub.it], is likely to arise naturally in this problem and is modeled as having two possible sources. First, while the production of research requires generally a continuous effort on the part of the researcher, observed publications are inherently discrete. Thus, publication measures for a given time interval may overstate the true output in the period (if the author had a backlog of articles accepted), or understate the true output (if the author had a rash of rejections). Moreover, any criterion used to compare research output across individuals is imperfect and is likely, for example, to exclude some relevant research outlets and give an inappropriate weight to some others that are included. In this case, observed output is measured with error that varies by individual and time, which we denote as [u.sub.it], where [u.sub.it] [sim] N(O,[[[sigma].sup.2].sub.u]).

Second, there may be idiosyncratic publishing tendencies of a given individual that are not fully captured by even a detailed list of explanatory variables. This unobserved publishing tendency is likely to be correlated over time; for simplicity, we specify the error process as AR(1), [rho][[epsilon].sub.it]. It is assumed that [rho][[epsilon].sub.it] and [u.sub.it] are independent, which essentially requires that the individual's unmeasured ability is uncorrelated with the measurement error in the dependent variable beyond that attributable to the first-order correlation. [12]

Thus, in the most general form, the error term can be expressed as:

(3) [[epsilon].sub.it] = [rho][[epsilon].sub.it-1] + [u.sub.it].

Our empirical analysis uses two different assumptions regarding the error. First, we estimate a standard tobit model that assumes that the only source of the randomness is the stochastic nature of the publication process, [u.sub.it] Second, we estimate an autoregressive tobit model using the error specification in equation (3), which is estimated using the hermite integral approximation of the type described by Butler and Moffit [1982]. [13] Comparisons of the two models provide some insights into the importance of unobserved heterogeneity in determining the assimilation and cohort effects.

IV. EMPIRICAL RESULTS

To make our analysis directly comparable to prior immigration studies, we restrict our analysis to Ph.D. economists working in the United States. The analysis first compares and contrasts the results of the cross-sectional and longitudinal models to examine whether assimilation and cohort effects that are found for broader populations of immigrants are also present for foreign-born economists. Subsequent analysis examines whether observed individual heterogeneity can explain the possible differential return to U.S. experience and cohort effects. Each of the specifications is estimated using both a standard and an autoregressive tobit model to examine the role of unobserved heterogeneity in assimilation and cohort quality.

Cross-Sectional and Longitudinal Estimates

The results applying cross-sectional and longitudinal specifications to the number of published pages using both the standard and autoregressive tobit models are presented in Table II. Each of specifications presented are replicated and are robust to the use of alternative publication measures. [14] Moreover, the coefficients on the human capital, gender, and period variables are also qualitatively equivalent across the models. In particular, research output appears to increase with experience, and publication skills depreciate with age. [15] Overall, the period effects suggest that research production has increased over time, which likely reflects both the increase in the number of research outlets and the expansion of institutions that require research for promotion. [16] In addition, female AEA members appear to produce significantly less than their male counterparts. Given the robustness of the findings for the non-immigrant-specific results, the remaining discussion focuses on the predicted differences b etween foreign- and native-born economists.

The first two columns of Table II replicate the traditional cross-sectional model of immigrant productivity and confirm prior findings that productivity increases at a decreasing rate with years since entry. Finding an assimilation effect from the first year of U.S. employment is not necessarily expected because 35% of foreign-born economists obtain their B.A. in the United States, and roughly 90% have a U.S. Ph.D. Thus the assimilation process is likely to have already begun before our defined point of U.S. labor-force entry. Moreover, the educational backgrounds of the foreign born who ultimately select a U.S. professional career are likely to have begun the assimilation process well before they actually migrate. Therefore, it is not surprising that the inclusion of controls for possible cohort effects (columns 3 and 4) and the use of panel techniques (columns 7 and 8) eliminate the significant differential return to U.S. experience for the foreign-born economists by controlling for observed and unobserved heterogeneity. To the extent that foreign-born economists are not expected to benefit from assimilation in these data, the finding of a significant assimilation effect in the cross-sectional specification suggests that the failure to control for changes in immigrant quality can yield predicted increases in productivity with years since entry when none is, in fact, present.

The cohort effects presented in Table II suggest a U-shaped productivity differential between native- and foreign-born economists from 1936 to 1985. This finding is similar to that found in Funkhouser and Trejo [1995], which includes a broader population of immigrants. In particular, the standard tobit model in columns 3 and 4 suggest that foreign-born economists entering the U.S. economics profession before 1941 and after 1975 produce significantly more research than their native-born counterparts, while those entering in the intervening years do not differ significantly from native-born economists. The greater research productivity of the foreign born prior to 1941 likely reflects the influx of highly skilled economists that migrated as a result of World War II. The post-1975 productivity differential may reflect both the 1972 immigration law changes that affected the ability of the foreign born to remain in the United States after their Ph.D. and the recession in academia beginning in the 1970s as document ed in Cartter [1976], Breneman [1975], and Brook and Marshall [1974]. Jointly, these changes might make economics departments relatively more selective in their hiring of foreign-born economists. [17]

The results for the autoregressive tobit model presented in columns 7 and 8 of Table II also suggest a U-shaped pattern for the cohort effects, but these results also indicate that the foreign-born are relatively more productive than the native born between 1956 and 1970. The findings concerning these latter immigrant cohorts may reflect other changes in immigration law that made it easier to migrate from Eastern Europe in the mid-1950s and Asia in the early 1960s. [18] These results suggest that the highly skilled from these countries may have had pent-up demand to migrate to the United States, which lead to a large, temporary influx of relatively able economists from these regions. Overall, these panel results strengthen the contention that immigrant quality changes over time in response to changes in immigration law. Nonetheless, while the cohort effects for the highly skilled roughly correspond to those of the broader population of immigrants, the results do not suggest a pronounced change in the quality of the foreign born corresponding to the 1965 Immigration Act.

Country Effects

The results presented in Table III introduce source-region controls and observed quality controls to the longitudinal specification. The results from the standard tobit model indicate that foreign-born economists from Eastern European nations produce significantly more research than native and Canadian-born economists. The autoregressive panel model that includes country controls indicates that the foreign-born from Asia, Africa, and the Caribbean publish significantly less than Canadian economists working in the United States. However, once the observed quality controls are applied, only foreign-born economists from Africa appear to be significantly different. This suggests that the source-region controls measure observed attributes of the foreign born that originate from the region. On the other hand, unlike LaLonde and Topel [1992] who suggest that much of the change in cohort quality can be attributed changes in source country, the magnitude and significance of the cohort effects do not change qualitativ ely when the binary variables for source region are used in the model. Moreover, most of the cohort effects do not change qualitatively when a detailed list of field, occupation, and quality of Ph.D. institution are introduced. Thus, the cohort effects appears to be relatively more robust for Ph.D. economists than for the broader population of immigrants.

V. CONCLUSION

This article uses a unique, unbalanced panel of native- and foreign-born members of the AEA in 1964, 1974, 1985, and 1989 to analyze how the research productivity of foreign-born economists has changed before and after the Immigration and Nationality Act of 1965 and with respect to their native-born counterparts. Our study provides some of the first formal evidence that the post-1965 shift in source-country mix of the broader population of U.S. immigrants away from Europe toward Asia also has occurred among the highly skilled. The analysis compares and contrasts cross-sectional, longitudinal, and panel specifications to examine the role of observed and unobserved heterogeneity in immigrant assimilation and cohort differences in productivity over time.

The empirical analysis first estimates a traditional cross-sectional specification and finds an immigrant-specific return to experience, which suggests that foreign-born economists assimilate over time. Our analysis uses a longitudinal specification to control for cohort effects, which eliminates the observed differential return to U.S. experience. Given that the assimilation process for foreign-born economists likely begins well before their first post-Ph.D. U.S. employment, this result is not surprising.

Consistent with the findings for broader populations of immigrants, the longitudinal specifications suggest a U-shaped pattern for the cohort effects. In particular, foreign-born economists are found to produce more research than their native-born counterparts prior to World War II, when the tyranny in Europe induced the highly skilled to migrate to the United States, and after 1975, when stricter visa requirements and a poor job market for academics may have made economics departments relatively selective. These cohort effects are robust to the inclusion of a detailed list of qualitative and regional variables. Moreover, with the inclusion of panel controls for unobserved heterogeneity, there is evidence to suggest that the relaxation of immigration restrictions on migrants from Eastern Europe in the 1950s and Asia in the 1960s also led to an influx of relatively productive economists. The foreign-born economists who entered the United States in the years immediately following World War II were less product ive than those who entered right before the war. However, the 1965 Immigration Act and its resulting change in the source-country compositional mix do not appear to have resulted in any decline in quality. Thus, overall, our findings suggest that foreign-born economists are at least as productive as their native-born counterparts and that the theoretically postulated brain drain of relatively able foreign-born economists to the United States did occur at several times during the postwar period.

(*.) We would like to thank George Borjas, Daniel Hamermesh, and Michael Greenwood for their insights and suggestions. We are responsible for any remaining mistakes.

McDowell: Professor, Department of Economics, Arizona State University, Tempe, Ariz. 85287-3806. Phone 1-480-965-7109, Fax 1-480-965-0748, E-mail john.mcdowell@asu.edu.

Singell: Associate Professor, Department of Economics, University of Oregon, Eugene, Or. 97403-1285. Phone 1-541-346-4672, Fax 1-541-346-1243, E-mail lsingell@oregon.uoregon.edu.

(1.) The biographical listings in the published ABA directories provide data on the year and place of birth in the 1964, 1969, and 1974 editions. After 1974, only the year of birth is available in the published AEA directories. For surveys after 1974, the computer tapes containing the responses to the AEA Survey of Members can be used to retrieve information on birthplace. Diamond and Haurin [1994] were the first to use the computer tapes for research purposes.

(2.) An AEA member is included if he or she is age 65 or less. For example, an economist who is 45 in 1964, 55 in 1974, and 66 in 1985 would be included in the panel for 1964 and 1974 but not 1985. Not all AEA members respond to the portion of the survey that relates to nativity. The percentages of respondents providing data on birthplace are 73.1% (1964), 68.1% (1974), and 71.1% (1985).

(3.) To a much lesser extent, other data sources also are used (e.g., American Men and Women of Sciences: Social and Behavioral Science).

(4.) The 1985 computer tape only provides information on whether the individual is native or foreign born. For foreign-born individuals, several steps were taken to identify the precise country of birth (or origin). First, if the individual is also found in the 1964, 1969, or 1974 published AEA directories, the country of birth is determined from these sources. For the remaining individuals who had degrees from foreign institutions (e.g., a U.S. Ph.D. but a B.A. from a foreign institution), the location of the foreign degree-granting institution is used as the country of origin. Finally, for the foreign born who had degrees only from U.S. institutions, a phone survey was conducted in which the individual was asked the place of birth. This survey resulted in country-of-birth data on 132 of 154 individuals surveyed. The 22 individuals for which precise country of birth could not be determined were dropped from the sample.

(5.) To keep a similar time span between cross-sections, a 1989 observation is included only if the individual is found in the 1993 AEA directory.

(6.) Only articles published in journals are enumerated in the publication counts. Thus, published books and articles published in books of collected works are not included.

(7.) Following Sauer's [1988] findings for the rewards to coauthorship, weighted measures use the simple 1/n rule. For "quality" of journal placement, the weighted measures use as weights the journal's "impact factor" as reported in the SSCI Journal Citation Reports. Native-and foreign-born individuals in the sample published articles in 412 publication outlets (as recorded in the Index of Economic Articles). A quality index or impact factor is available for 224 of these 412 publication outlets. For the remaining 188 outlets, a weight of zero is used. Of the 188 outlets assigned a weight of zero, 106 outlets are not contained in the World List of Social Science Periodicals [UNESCO 1991], which lists 4,459 periodicals that meet their criteria of being a "scientific periodical." Only 11 of the remaining 82 journals, which are noted as a "scientific periodical" in the World List but are assigned a zero weight, are published in the United States.

(8.) Typically, these analyses track the progress of a particular cohort over successive cross-sectional data bases. Analyses using actual longitudinal data are provided in Chiswick [1978b], Jasso and Rosenzweig [1986, 1988], and Borjas [1989].

(9.) McDowell and Singell [1998] provide a more detailed, descriptive analysis of these data.

(10.) Academics are distinguished from other occupations by three binary variables that are equal to one if the individual is employed in business, government, or other nonacademic institutions. Two schooling measures are constructed from departmental publication rankings in Graves et al. [1982]. First, a binary variable that equals one if the individual received a Ph.D. from a top 35 program is used to broadly distinguish research- versus nonresearch-oriented departments. Second, this variable is interacted with the log of the Graves ranking to distinguish among research-oriented institutions. Because foreign institutions are not included in the Graves ranking, a binary variable that is equal to one for those individuals who obtain a Ph.D. from an institution located outside the United States is also included. Immigrants who received their B.A. in the United States are also distinguished from those who did not, because migration decisions that occur at a younger age (particularly if the migrant is a child) a re likely to differ from those that occur for employment reasons. Finally, a set of 15 dummy variables is included to control for the primary area of research interest, where the category "general" (as defined by Journal of Economic Literature prior to 1991) is excluded.

(11.) The period-effect restriction is the minimum identification requirement. The human-capital and personal-attribute coefficients are also restricted to be the same for native- and foreign-born economists because less restrictive empirical specifications find no statistical differences by origin of birth. Thus, we follow prior work and focus on cohort and year-since-entry effects.

(12.) As an alternative to the proposed model, the random-effects tobit model specifies an individual-specific random component, [[alpha].sub.i], that measures unobserved heterogeneity instead of [rho][[epsilon].sub.it]. Hsiao [1986] proposes including a lag of research output and conducting a t-test on its coefficient to confirm the possible presence of an individual effect. This test is conducted for several specifications and supports the presence of an individual effect, which could arise from several sources, including serial correlation or unobserved heterogeneity. We select the AR(1) model because its likelihood function is significantly more stable in convergence; replication of our AR(1) specifications using a random-effects tobit model yield the same qualitative conclusions when the random-effects model converges.

(13.) A first-order, autoregressive structure can arise if there is serial correlation or state dependence. The difference between the two hypothesis is that the former implies research output depend on the entire history of the errors, whereas the latter output depends only on a one-period lag of output. We conduct a test for state dependence proposed by Chamberlain [1982] that uses a likelihood-ratio test to compare a specification with and without the lagged values of the time varying explanatory variables. The chi-squared statistic of 14.1 rejects the null of state dependence at the 5% level. Thus, the AR(1) model is used in the estimations.

(14.) As a point of comparison, columns 1 and 2 of Appendix Table Al include the base longitudinal specification using the number of articles weighted by coauthors and the quality index, which is the publication measure that is most distinct from published pages. The results are qualitatively equivalent but yield less significant cohort effects than those presented subsequently. The descriptive statistics provided by McDowell and Singell [1998] indicate that foreign-born economists are more likely to coauthor than native-born economists, particularly those in early cohorts, which may account for the smaller observed cohort effect when using a coauthor adjusted publication measure.

(15.) The negative coefficient on age could also reflect unobserved differences in ability or motivation between those who finish their Ph.D. at older and younger ages.

(16.) Although the number of published pages have increased over time, the number of articles weighted by quality and coauthors has not increased (Appendix Table A1).

(17.) Native-born economists have been significantly more likely to work in nonacademic positions than their foreign-born counterparts, particularly in recent years. However, two longitudinal specifications using only academics presented in columns 3 and 4 of Appendix Table A1 yield the same qualitative conclusions as those that use all AEA members. Thus, the cohort effects do not appear to be related to different occupational choices by native- and foreign-born economists.

(18.) For instance, during the 1950s, Congress passed several acts that facilitated the flow of refugees (e.g., Eastern Europeans) to the United States. In addition, the Act of September 26, 1961, eliminated the ceiling of 2,000 on the aggregate quota of the Asia-Pacific triangle, thus partially lowering restrictions that had been particularly binding on potential immigrants from Asian nations.

REFERENCES

Bhagwati, Jagdish, and Carlos Rodriquez. "Welfare-Theoretical Analysis of the Brain Drain." Journal of Development Economics, September 1975, 193-221.

Borjas, George J. "Assimilation, Changes in Cohort Quality, and the Earnings of Immigrants." Journal of Labor Economics, October 1985, 463-89.

-----. "Self-Selection and the Earnings of Immigrants." American Economic Review, September 1987, 531-53.

-----. "Immigrant and Emigrant Earnings: A Longitudinal Study." Economic Inquiry, January 1989, 21-37.

-----. Friends or Strangers: The Impact of Immigration of the U.S. Economy. New York: Basic Books, 1990.

-----. "National Origin and the Skills of Immigrants in the Postwar Period," in Immigration and the Work Force: Economic Consequences for the United States and Source Areas, edited by George J. Borjas and Richard B. Freeman. Chicago: University of Chicago Press, 1992, 17-47.

-----. "Assimilation and Changes in Cohort Quality Revisited: What Happened to Immigrant Earnings in the 1980s?" Journal of Labor Economics, April 1995, 201-45.

Breneman, David W. Graduated School Adjustments to the "New Depression" in Higher Education. National Board on Graduate Education, Technical Report Number Three, February 1975.

Brook, Kathleen, and J. Ray Marshall. "The Labor Market for Economists." American Economic Review, May 1974, 488-511.

Butler, J. S., and Robert Moffit. "A Computationally Efficient Quadrature Procedure for the One-Factor Mutinomial Probit Model." Econometrica, May 1982, 761-65.

Cartter, Allan M. Ph.D.'s and the Academic Labor Market. New York: McGraw-Hill, 1976.

Chamberlain, Gary. "Multivariate Regression Models for Panel Data." Journal of Econometrics, January 1982, 5-46.

Chiswick, Barry R. "The Effect of Americanization of the Earnings of Foreign-Born Men." Journal of Political Economy, October 1978a, 897-921.

-----. "A Longitudinal Analysis of the Occupational Mobility of Immigrants," in Industrial Relations Research Association, Proceedings of the Thirteenth Annual Winter Meetings, December 1977. New York. Madison, Wis.: IRRA, 1978b, 20-27.

-----. "Is the New Immigration Less Skilled Than the Old?" Journal of Labor Economics, April 1986, 168-92.

Diamond, Arthur M., Jr., and Donald R. Haurin. "Determinants among Ph.D. Economists of Membership in a Professional Association." Education Economics, 2(1), 1994, 13-28.

Funkhouser, Edward, and Stephen J. Trejo. "The Labor Market Skills of Recent Male Immigrants: Evidence from the Current Population Survey." Industrial and Labor Relations Review, July 1995, 792-811.

Graves, Philip E., James R. Marchand, and Randall Thompson. "Economics Department Rankings: Research Incentives, Constraints, and Efficiency." American Economic Review, December 1982, 1131-41.

Greenwood, Michael J. "The Economics of Mass Migration from Poor to Rich Countries: Leading Issues of Fact and Theory." American Economic Review, May 1983, 173-77.

Greenwood, Michael J., and John M. McDowell. "The Factor Market Consequences of U.S. Immigration." Journal of Economic Literature, December 1986, 1738-72.

Grubel, Herbert G., and Anthony Scott. "The Characteristics of Foreigners in the U.S. Economics Profession." American Economic Review, March 1967, 131-45.

Hansen, W. Lee, Burton A. Weisbrod, and Robert P. Strauss. "Modeling the Earnings and Research Productivity of Academic Economists," Journal of Political Economy, August 1978, 729-41.

Hsiao, Cheng. Analysis of Panel Data. Cambridge: Cambridge University Press, 1986.

Huang, Wei-Chiao. "A Pooled Cross-Section and Time-Series Study of Professional Indirect Immigration to the United States." Southern Economic Journal, July 1987, 95-109.

Jasso, Guillermina, and Mark R. Rosenzweig. "What's in a Name? Country-of-Origin Influences on the Earnings of Immigrants in the United States," in Human Capital and Development, vol. 5, edited by Oded Stark and Ismail Sirageld. Stamford, Conn.: JAI Press, 1986, 75-106.

-----. "How Well Do U.S. Immigrants Do? Vintage Effects, Emigration Selectivity, and Occupational Mobility," in Research in Population Economics, edited by T. Paul Schultz. Stamford, Conn.: JAI Press, 1988, 229-253.

LaLonde, Robert J., and Robert H. Topel. "Immigration in the American Labor Market: Quality, Assimilation, and Distributional Effects." American Economic Review, May 1991, 297-302.

-----. "Assimilation of Immigrants in the U.S. Labor Market," in Immigration and the Work Force: Economic Consequences for the United States and Source Areas, edited by George J. Borjas and Richard B. Freeman. Chicago: University of Chicago Press, 1992, 67-92.

McDowell, John M. "Obsolescence of Knowledge and Career Publication Profiles: Some Evidence of Differences among Fields in Costs of Interrupted Careers." American Economic Review, September 1982, 752-68.

McDowell, John M., and Larry D. Singell, Jr. "The Foreign Born in the U.S. Economics Profession: Attributes, Composition, and Relative Productivity." Working paper, University of Oregon, 1998.

Oster, Sharon M., and Daniel S. Hamermesh. "Aging and Productivity among Economists: What and Why?" The Review of Economics and Statistics, February 1998, 154-56.

Sauer, Raymond D. "Estimates of the Returns to Quality and Coauthorship in Economic Academia." Journal of Political Economy, August 1988, 855-66.

UNESCO. World List of Social Science Periodicals, 8th. ed. Paris: United Nations Educational, Scientific and Cultural Organization, 1991.