The NBER Reporter 2007 Number 4: Conferences
Inter-American Seminar on Economics
Economics of Agglomeration
The Economics of High-Skill Labor Markets
Although decades of empirical research has demonstrated that criminal behavior responds to incentives, non-economists frequently express the belief that human beings are not rational enough to make calculated decisions about the costs and benefits of engaging in crime therefore, they a priori draw the conclusion that criminal activity cannot be altered by incentives. However, scientific research should not be driven by personal beliefs. Whether economic conditions matter, or deterrence measures such as police, arrests, prison deaths, executions, and commutations provide signals to people is an empirical question, which should be guided by a solid theoretical framework. Mocan and Gittings examine the relationship between state homicide rates and death-penalty related outcomes (executions, commutations, and removals from death row). They alter econometric models in a number of directions to make the deterrence results disappear by deliberately deviating from the theoretically consistent measurement of the risk variables in a variety of ways. They further investigate the sensitivity of the results to changes in the estimation sample (for example by removing high executing states from the sample) and weighting. Their basic results are insensitive to these and a variety of other specification tests performed in the paper. The results are often strong enough to hold up even under theoretically meaningless measurements of the risk variables. The findings are robust, providing evidence that people indeed react to incentives induced by capital punishment.
After reaching a historic peak by the end of the 1990s, homicides in large cities in the state of São Paulo dropped sharply. Several explanations have been advanced, most prominently improvements in policing, adoption of policies such as dry laws, and increased incarceration. DeMello and his co-authors show that demographic changes play a large role in explaining the dynamics of homicide. More specifically, they present evidence of a strong co-movement between the proportion of males in the 15-25 age bracket and homicides at the statewide and at city levels, and argue that the relationship is causal. They estimate that a 1 percent increase in the proportion of 15-to-24-year-old males causes a 4.5 percent increase in homicides.
Between 2000 and 2006 the murder rate in Newark doubled while the national rate remained essentially constant. Newark now has eight times as many murders per capita as the nation as a whole. Furthermore, the increase in murders came about through an increase in lethality: total gun discharges rose much more slowly than the likelihood of death per shooting. In order to explain these trends, O'Flaherty and Sethi develop a theoretical model of murder in which preemptive killing and weapon choice play a central role. Strategic complementarity amplifies changes in fundamentals, so areas with high murder rates (war zones) respond much more strongly to changes in fundamentals than those with low murder rates (peaceable kingdoms). In Newark, the changes in fundamentals that set off the spiral were a collapsing arrest rate (and probably a falling conviction rate), a reduction in prisoners, and a shrinking police force.
Internal conflicts entail large asset losses for segments of the civil population. Asset losses may compromise future welfare of households, leaving a legacy of structural poverty that is difficult to overcome. Ibanez and Moya analyze how asset losses occur during internal conflicts and the process of asset recovery after the shock. In order to achieve this objective, they concentrate on a particularly vulnerable group of victims of war: the displaced population in Colombia. They adopt quantitative and qualitative approaches to achieve this objective by: providing a detailed description of losses stemming from forced displacement; analyzing qualitative evidence to understand asset recovery processes for the displaced population; and estimating OLS, Instrumental Variable, and quartile regressions to identify the determinants of asset recovery. Their results indicate that recuperating or accumulating new assets is a rare event: only 25 percent of households are able to recuperate assets, and asset ownership seems still insufficient to overcome poverty. In addition, displaced households do not catch up as settlement in destination sites consolidates so, unless a positive intervention is implemented, displaced households will be locked in a low income trajectory and leaping forward to a high return asset level is highly unlikely.
Domestic violence remains a major public policy concern despite two decades of policy intervention. To eliminate police inaction in response to domestic violence, many states have passed mandatory arrest laws, which require the police to arrest abusers when a domestic violence incident is reported. These laws were justified by a randomized experiment in Minnesota that found that arrests reduced future violence. This experiment was conducted during a time period when arrest was optional. Using the FBI Supplementary Homicide Reports, Iyengar finds that mandatory arrest laws actually increased intimate partner homicides. She hypothesizes that this increase in homicides is attributable to decreased reporting. She investigates the validity of this reporting hypothesis by examining the effect of mandatory arrest laws on family homicides where the victim is less often responsible for reporting. For family homicides, mandatory arrest laws appear to reduce the number of homicides. This study therefore provides evidence that these laws may have perverse effects on intimate partner violence, harming the very people they seek to help.
Schargrodsky and Di Tella study the arrest rates of individuals released from prison as opposed to those who were formerly being electronically monitored. They exploit the fact that electronic monitoring in Argentina is assigned to prisoners by judges with highly different preferences: some regularly assign prisoners to electronic monitoring while others never do so. Importantly, electronic monitoring is assigned without the requisite of a concomitant rehabilitation plan, facilitating the discussion of electronic monitoring as an individual policy. The researchers find that those "treated" with electronic monitoring have marginally lower arrest rates following their release than former prisoners treated in standard jails. Offenders who went through the electronic monitoring program have approximately 8 percent lower recidivism than offenders who went to prison. Moreover, among inmates under electronic monitoring, recidivism rates are actually lowest for those who by-passed jail by going directly from the police station to electronic surveillance. Offenders with a previous criminal record probably should be excluded from the electronic monitoring system, as their evasion and recidivism rates are high.
Dills, Summers, and Miron evaluate what economists have learned over the past 40 years about the determinants of crime. They base their evaluation on two kinds of evidence: an examination of aggregate data over long time periods and across countries, and a critical review of the literature. They argue that economists know little about the empirically relevant determinants of crime. This conclusion applies both to policy variables like arrest rates or capital punishment and to less conventional factors, such as abortion or gun laws. The reason is that even hypotheses that find some support in U.S. data for recent decades are inconsistent with data over longer horizons or across countries. The hypothesis that drug prohibition generates crime, however, is consistent with the long times-series and the cross-country facts about crime.
Medina and his co-authors study crime in Bogotá and find that households living in the highest socioeconomic stratum, stratum 6, are paying up to 7.2 percent of their house values to keep their average homicide rates the same and to avoid increasing them by a single standard deviation. Households in strata 5, those in the next richest group in the city, would be paying up to 2.4 percent of their house values to keep their average homicide rates. These results reveal the willingness to pay for security by households in Bogotá, and additionally, reveal that a supposed pure public good, like security, ends up propitiating urban private markets that auction security. These markets imply different levels of access to public goods among the population, and actually, the exclusion of the poorest.
Draca, Machin, and Witt contribute to two relatively small, but growing, literatures of high public policy relevance on the causal impact of police on crime and the economics of terrorism. They bring the two together by looking at what happened to crime before and after the terror attacks that hit central London in July 2005. The attacks resulted in a large redeployment of police officers to central London boroughs as compared to outer London in fact, police deployment in central London increased by over 30 percent in the six weeks following the July 7 bombings. In this time period, crime fell significantly in central relative to outer London. Study of the timing of the crime reductions and their magnitude, which types of crime were more affected, and a series of robustness tests looking at possible biases make them confident that their research approach identifies a causal impact of police on crime. Implementing an instrumental variable approach shows an elasticity of crime with respect to police of approximately -0.3, so that a 10 percent increase in police activity reduces crime by around 3 percent.
Using census data for Argentine prisons for the period 2002-2005, Alzua, Rodriguez, and Villa present evidence of the positive effect that prisoner education programs have on in-prison conflictivity, measured as sanctions or violent behavior of the prisoner. In order to overcome the problems of endogeneity of education, they explore an instrumental variables as well as a panel fixed effects approach. Their results show a decrease in participation in violent conflicts and bad behavior, which can be partially attributed to education.
The control of drug activity currently favors supply-side policies: drug suppliers in the United States face a much higher arrest rate and longer sentences than drug demanders. Chang, Coulson, and Wang construct a simple model of drug activity with search and entry frictions in labor employment and drug transactions, where the drug price and the distribution of population in a community are determined according to an occupational choice rule and a downward-sloping drug demand schedule. Through calibration analysis, they find a strong "dealer replacement effect." As a result, it is beneficial to lower the supply arrest rate and to raise the demand arrest rate from their current values. A tax-revenue neutral shift of 10 percent from supply-side to demand-side arrests can reduce the population of potential drug dealers by 22-25 thousand and raise aggregate local income by 380-400 million dollars at 2002 prices.
Soares and Naritomi discuss the pattern, causes, and consequences of the high crime rates observed in Latin America. Crime represents a substantial welfare loss and a potentially serious hindrance to growth. They conduct an informal assessment of the relative strength of the alternative hypothesis raised in the literature to explain the phenomenon. They argue that, despite being extremely high, the incidence of crime in the region is not much different from what should be expected based on socioeconomic and public policy characteristics of its countries. Estimates from the empirical literature suggest that most of its seemingly excessively high violence can be explained by three factors: high inequality, low incarceration rates, and small police forces. Still, country specific experiences have been different in many respects. The evidence suggests that effective policies toward violence reduction do exist and have been shown to work within the context of Latin America itself.
It is anticipated that these papers and discussions will be published in a University of Chicago Press conference volume. Its availability will be announced in a future issue of the NBER Reporter. The papers will also be available at "Books in Progress" on NBER's website.
One of the great attractions of cities is that urban proximity speeds the flow of ideas. Improvements in communications technology could erode this advantage and allow people and firms to decentralize. On the other hand, if cities have an edge in producing new ideas, then communication technology may strengthen the demand for cities by increasing the returns to innovation. Improvements in communication technology can increase the returns to innovation by allowing new ideas to be used in more geographic locales. Glaeser and Ponzetto present a model that illustrates these two rival effects of communication technology on cities. They then show that the model can help us to understand why the past 35 years have been kind to some cities, like New York and Boston, and devastating to others, like Cleveland and Detroit.
The empirical finding of Zipf's law for the size distribution of cities is one of the most celebrated in all of urban economics. In the literature, size generally refers to total population; this can be bigger in one city than another, both because of higher population density and because of having more land area. But one problem is that the land area of city units is often defined by arbitrary political and administrative considerations as opposed to considerations that urban economists would use to define the boundaries for scientific study. Holmes and Lee therefore examine the size distribution of cities looking only at the density margin, holding fixed the land-area margin. They focus on a particular land unit, a six-by-six mile square grid. Most of the United States was drawn up into such grids in the early 1800s for the purpose of land sales, and the resulting six-by-six mile squares were called townships. The researchers examine the population distribution across townships. Their main finding is that Zipf's law does not hold with six-by-six mile squares. When they do a Zipf's plot, not only is the slope not one, it is not even a straight line. Nevertheless, they do find some interesting empirical regularities that are robust over different cuts of the data. For example, the Zipf's plot has three regions. For small cities, the distribution is certainly not Pareto where there are always more cities, the smaller the city size. Instead, on this end of the distribution, the density of city sizes actually decreases as size decreases. For intermediate size cities, the distribution is Pareto with a slope around 0.8. For large cities, the distribution is Pareto with a slope around 2. The researchers also propose an explanation for this "kink" between the intermediate and large cities using the monocentric city model.
The ethnic composition of U.S. inventors is undergoing a significant transformation with deep impacts for the overall agglomeration of U.S. innovation. Kerr applies an ethnic-name database to individual U.S. patent records to explore these trends in greater detail. He observes that the contributions of Chinese and Indian scientists and engineers to U.S. technology formation increase dramatically in the 1990s. At the same time, these ethnic inventors became more spatially concentrated across U.S. cities. The combination of these two factors helps to stop and reverse long-term declines in overall inventor agglomeration evident in the 1970s and 1980s. The heightened ethnic agglomeration is particularly evident in industry patents for high-tech sectors, and no similar trends are found in institutions constrained from agglomerating (for example, universities, or government).
Because of fixed costs, additional people nearby confer a benefit on each other by helping to make more retail products available. Yet, because product preferences differ across groups of consumers, the appeal of what's available depends on the mix of consumers. If product preferences relate to consumer characteristics such as race, income, age, and ethnicity, then product availability will be stimulated by concentration of like individuals. The sensitivity of product availability to demographic mix of consumers has been documented for metropolitan-area products, such as newspapers, radio, and television, as well as one neighborhood-level product, restaurants. Waldfogel revisits the question for a broader group of local retail establishments. Using the Consumer Expenditure Survey, he documents that preferences differ across groups. Then, using the 2000 Census and the 2000 Zipcode Business Patterns, he shows that the mix of products available is sensitive to the mix of local preferences. P therefore derive benefit through the product market from agglomerating with persons with similar product preferences, and this may help to explain patterns of residential segregation.
There is substantial variation in the quality of care delivered by different hospitals, and while overall quality has been rising over time, many hospitals still lag behind. Baicker and Chandra explore potential sources of this variation, and examine whether hospitals "learn" to provide higher quality from their neighbors, particularly in cities. They examine detailed hospital-level data on adherence to best practices in the treatment of heart attack, heart failure, and pneumonia for 2004-6, along with information on population and hospital characteristics. They find that hospitals are likely to improve the quality of care that they offer more quickly if there are nearby hospitals offering superior care. Hospitals are particularly likely to learn from nearby teaching hospitals, and while city hospitals often begin by offering lower quality care, they are likely to converge towards the standard of care offered by their higher quality neighbors.
Rosenthal and Strange consider the relationship between local industrial organization and agglomeration economies. They begin by presenting a model of agglomeration, industrial organization, and entrepreneurship. The model's key prediction is the existence of a virtuous circle of urban entrepreneurship: the presence of small establishments produces an environment conducive to growth, in particular entrepreneurial growth. They then investigate this prediction empirically, showing that additional activity at smaller establishments is associated with a larger amount of entrepreneurial activity.
Overman and Puga provide empirical evidence on the role of labor market pooling in determining the spatial concentration of U.K. manufacturing establishments. This role arises because large concentrations of employment iron out idiosyncratic shocks and improve establishments's ability to adapt their employment to good and bad times. The researchers measure the likely importance of labor pooling by calculating the fluctuations in employment of individual establishments relative to their sector and averaging by sector.Their results show that sectors whose establishments experience more idiosyncratic volatility are more spatially concentrated, even after controlling for a range of other industry characteristics that include a novel measure of the importance of localized intermediate suppliers.
Crime, pollution, and congestion pose major challenges to big-city quality of life. Kahn uses several datasets to examine recent trends. In the case of pollution and crime, big cities recently have experienced sharp improvements. Still, commute times continue to be longer in big cities relative to smaller cities. Kahn examines the role of suburbanization of jobs and households in explaining these observed patterns. Employment suburbanization has reduced commute times for suburban residents. Residential suburbanization also has reduced population exposure to urban crime and pollution. These two disamenities are spatially concentrated in the center cities. This paper highlights the quality of life benefits from population sprawl.
Combes and his co-authors revisit the estimation of urbanization economies using French wage and TFP data. To deal with the "endogenous quantity of labor" bias (that is, urban agglomeration as a consequence of high local productivity, rather than a cause), they take an instrumental variable approach and introduce a new set of geological instruments, in addition to standard historical instruments. To deal with the "endogenous quality of labor" bias (that is, cities attract skilled workers so that the effects of skills and urban agglomeration are confounded), they take a fixed effect approach with wage data. They find modest evidence about the endogenous quantity of labor bias and both sets of instruments give a similar answer. They find that the endogenous quality of labor bias is quantitatively more important.
Urban success increasingly has taken two different forms in the post-war era. One involves very high house price growth with relatively little population growth, while the other pairs strong population expansion with mild house price appreciation. Associated with this is a growing spatial dispersion of house prices and incomes, both across and within markets. Gyourko and his co-authors document the nature of this dispersion; they then ask what can account for this relationship, and propose and analyze two basic explanations. One is differences in urban productivity attributable to agglomeration economiesa mainstay of urban economics. The other is a preference-based explanation that relies on sorting, with the rich ultimately outbidding others for the scarce slots available in supply-constrained metropolitan areas. The evidence suggests that pure sorting is at least partially responsible for the urban outcomes they see, but it also is clear that much more work is needed to pin down the relative contributions of these two basic factors.
Economic research on industry location and agglomeration has focused nearly exclusively on manufacturing. Kolko shows that services are prominent among the most agglomerated industries, especially at the county level. Because traditional measures of knowledge spillovers, natural resource inputs, and labor pooling explain little of agglomeration in services industries, he takes an alternative approach and looks at co-agglomeration to assess why industries cluster together. By considering the location patterns of pairs of industries instead of individual industries, the traditional agglomeration explanations can be measured more richly, and additional measures like the need to locate near suppliers or customers can be incorporated. The results show that co-agglomeration between pairs of services industries is driven by knowledge spillovers and the direct trading relationship between the industries, especially at the zip code level. Information technology weakens the need for services industries to co-agglomerate at the state level, perhaps because electronic transport of services outputs lowers the value of longer-distance proximity. These results are in sharp contrast to results for manufacturing, for which labor pooling contributes most to co-agglomeration, and the direct-trading relationship contributes more to state-level co-agglomeration. These differences between services and manufacturing are consistent with simple models of transport costs.
These conference proceedings will be published by the University of Chicago Press in an NBER Conference Volume. Its availability will be announced in the NBER Reporter. The papers will also be available at "Books in Progress" on the NBER's website.
Elfenbein and his co-authors examine the determinants of transitions from paid employment into self employment by scientists and engineers between 1995 and 2001. They find that those working in small firms are significantly more likely to become self employed than those working in large firms. Entrepreneurs coming from both small firms and from large firms are more likely to be high performers (stars) or low performers (slugs) as measured by their pay in their prior jobs; this finding is particularly pronounced among those leaving small firms. Finally, there is some evidence that entrepreneurs coming from small firms perform better than those who come from large firms: they are more likely to persist in self-employment and earn more in their first period of self-employment, controlling for a number of other factors. The researchers explore the degree to which these relationships can be explained by theories that focus on differences in opportunity costs, the strength of pay-performance relationships, and diversity of activity sets, and the acquisition of entrepreneurial capital in firms of different size.
Hussey investigates the labor market effects of obtaining an MBA degree. Despite estimated positive returns to the degree on earnings, there is a negative reduced-form effect of the MBA on managerial attainment, a finding that is especially puzzling since managerial status and earnings are highly positively correlated. This reduced-form result is extremely robust. In particular, the evidence shows that the negative effect of an MBA on managerial attainment is not attributable to: observed or unobserved productivity differences between MBAs and non-MBAs; differential worker sorting across observable dimensions like firm size or industry; or a probationary period following MBA attainment. While basic models of MBA attainment cannot explain the key empirical regularities, Hussey's model of worker sorting on the basis of differential preferences over unobserved job or career tracks does allow for these findings. The model is estimated using a panel data set consisting of registrants for the Graduate Management Admissions Test (GMAT). Estimation produces a positive overall effect (on wages and managerial attainment) of an MBA, and demonstrates the existence of a compensating differential between positions with a higher likelihood of managerial attainment versus positions with a lower likelihood of managerial attainment.
The right to an equal and fair trial regardless of wealth is a hallmark of American jurisprudence. To ensure this right, the government pays attorneys to represent financially needy clients. In the U.S. federal court system, indigent defendants are represented by either public defenders who are salaried employees of the court or private attorneys, known as Criminal Justice Act (CJA) attorneys, who are compensated on an hourly basis. Iyengar measures differences in performance of these types of attorneys and explores some potential causes for these differences. Exploiting the use of random case assignment between the two types of attorneys, her analysis of federal criminal case level data from 1997-2001 from 51 districts indicates that public defenders perform significantly better than CJA panel attorneys in terms of lower conviction rates and sentence lengths. An analysis of data from three districts linking attorney experience, wages, law school quality, and average caseload suggests that these variables account for over half of the overall difference in performance. These systematic differences in performance disproportionately affect minority and immigrant communities and as such may constitute a civil rights violation under Title VI of the Civil Rights Act.
Ginther and Kahn first address whether academic women are actually less likely than men to choose to change employers for a better job. Indeed, there are strikingly different rates of mobility for those not forced to move those with tenure compared to those who are untenured (but tenure-track) or non-tenure track. So, they concentrate this examination of mobility on tenured academics who are not forced to leave, but instead only leave voluntarily. They construct two different measures of a "better job." One measure of advancement is a higher salary in the new job; their second measure of a better job combines institutional prestige and the person's academic rank into a single "academic job quality" measure. They find that tenured women are more likely than tenured men to move to better jobs, particularly those with higher ranks rather than higher salaries. Children hurt men's mobility to better jobs but actually may help women's. Marriage does not affect either tenured men's or women's mobility. Finally, of those who move to higher paying jobs, there is no evidence that "women don't ask," except at the very highest levels.
Where does adaptation to innovation take place? The supply of educated workers and local industry structure matter for the subsequent location of new work, that is, new types of labor-market activities that closely follow innovation. Using census 2000 microdata, Lin shows that regions with more college graduates and a more diverse industrial base in 1990 are more likely to attract these new activities. Across metropolitan areas, initial college share and industrial diversity account for 50 percent and 20 percent, respectively, of the variation in selection into new work unexplained by worker characteristics. Lin uses a novel measure of innovation output based on new activities identified in decennial revisions to the U.S. occupation classification system. New work follows innovation, but unlike patents, it also represents subsequent adaptations by production and labor to new technologies. Further, workers in new activities are more skilled, consistent with skill-biased technical change.
Cappelli and Hamori examine an important measure of job search in the context of executive-level jobs, using a unique dataset from a prominent executive search firm that identifies whether executives have pursued offers to be considered for a position at other companies. The unique attribute of this search process is that the initial call from the search firm is essentially an exogenous event for the executives: they also learn so little about a possible offer at that point that their decision has to rely on a judgment about the likelihood that an eventual offer is better than their reservation price. The researchers find support for the commonsense notion that individuals whose current job circumstances appear superior to those elsewhere are less inclined to search. They also find, as search theory predicts, that executives with longer experience with a firm are less likely to search. Finally, they find that the broader an executive's job experience across different labor market segments, the more likely they are to search, suggesting that a portfolio of experiences makes the likelihood of a good match with an unknown job offer more likely.
Oreopoulos and his co-authors analyze the long-term effects of graduating in a recession on earnings, job mobility, and employer characteristics for a large sample of Canadian college graduates with different predicted earnings using matched university-employer-employee data from 1982 to 1999. They then use their results to assess the role of search frictions in the labor market. They find that young graduates entering the labor market in a recession suffer significant initial earnings losses that eventually fade, but only after eight to ten years. They also document substantial heterogeneity in the costs of recessions and important effects on job mobility and employer characteristics, but small effects on time worked. They show that these adjustment patterns could be explained by endogenous search for better employers in the presence of time-varying search costs and comparative advantage. All results are robust to an extensive sensitivity analysis including controls for correlated business cycle shocks after labor market entry, endogenous timing of graduation, permanent cohort differences, and selective labor force participation.
Since the early 1990s the pay gap between men and women in the top five executive positions of S&P 1500 companies has persisted in the range of 25 to 45 pecent. One explanation is that executives who sit on the board, or "insiders," favor their male colleagues rather than rewarding performance when they set executive pay. Selody examines whether greater corporate board independence alleviates the gender gap in executive compensation. She tests the effect of the 2003 change to NYSE/NASD Corporate Governance Listing Standards that required boards to become more independent and disallowed inside board members to serve on the compensation committee. She compares the gender gap in compensation in firms that were required to convert to more independent boards with those that were not, before and after the reform. She finds that the gender gap in compensation increased in firms that converted to more independent boards relative to firms that were not required to increase board independence. She then considers two main hypotheses: 1) that more independent boards are more committed to shareholder value, and so pay women executives closer to their marginal product; or, 2) that outsiders on the board have less information about executives' marginal product and so initially rely on imprecise or biased prior beliefs about women's ability as a group until they learn more about individual executives' productivity. Additional evidence supports the second hypothesis. The increased gender pay gap in firms whose boards became more independent diminishes as board members learn more about executives' productivity through repeated observation. And, the increase in the gap is not uniform: it does not widen for CFOs and Legal Counsels, occupations with more specific credentials. Selody also finds an asymmetry in the response of men's and women's compensation to movements in market value. Women's compensation increases comparably to men's when market value increases, but decreases by more than men's when market value decreases, consistent with the hypothesis that boards' prior beliefs come into play when they attribute an outcome to luck or ability.
School principals increasingly are being called on to provide educational leadership in schools in the United States. Although prior studies have determined that individual principals can make a big difference to a school's trajectory, relatively little is known about their career profiles and the rewards for success. The labor market for principals potentially can inject a profit motive for improving public schools in what otherwise is considered to be a relatively noncompetitive market. Cullen and Mazzeo rely on individual longitudinal data on the universe of Texas principals from 1989 through 2006 to track how levels and changes in school performance affects principals' salaries and mobility across schools. They test whether the dramatic enhancements in school accountability that presumably increased principal accountability moderate these relationships.
Alter studies the gains to productivity from joint work of individuals with different levels of human capital within a firm. The analysis is based on a unique panel dataset that documents partners, investments, and investment outcomes of venture capital (VC) firms in California during 1979-2002. Alter finds that more successful venture capitalists recruit more new partners and that this effect increases when there is growth in the aggregate number of VC investments. Based on these findings, he develops and estimates a structural model that quantifies the increase in venture capitalists' productivity through skill leverage in the firm. He finds that joint work increased productivity of experienced and inexperienced partners by 8 percent.