Paper Session I:  Firms, Capital and the Economy

Joshua Linn, University of Illinois at Chicago

ABSTRACT:
There is considerable empirical evidence that oil prices had a large effect on the U.S. economy from World War II until the 1980s. Previous research has proposed several possible explanations, but has provided little direct evidence. In this paper I argue that linkages between manufacturing industries play an important role in amplifying oil price shocks. For example, consider an industry that uses energy and intermediate materials. An increase in the price of energy would cause output to contract because energy is an input. Furthermore, if the industry uses materials that require a lot of energy to be produced (e.g., paper), then the price of materials would rise. This indirect effect would cause a further decrease in output. On the other hand, there may be demand effects. Industries that supply their products to energy intensive industries would experience a large decrease in the demand for their output. For example, steel is an important input in aircraft manufacturing which is energy intensive. An oil price increase would reduce aircraft production, reducing the demand for steel and causing a decline in steel output.

In this paper I quantify the importance of three effects on value added. A priori, energy and materials cost shares should determine their magnitudes for a given industry. The first effect is the direct effect of the price of energy; value added should contract most in energy intensive industries. The second is the supply effect, which should be larger if an industry uses energy intensive products. The industry would experience a large increase in the prices of those energy intensive materials. The third is the demand effect, which should be larger if the industry’s output is used by energy intensive industries.

Using data from the Census of Manufactures from 1963-1997, I measure the relative importance of the three effects. I first construct measures of the demand and supply linkages at the four digit industry level from the materials files in the Census of Manufactures.

I then decompose changes in industry level value added into the direct, demand and supply effects. I find that the supply effect is considerably more important than the demand effect in explaining variation in industry level value added. This result is consistent with the fact that the average manufacturing industry uses large amounts of energy intensive products. On the other hand, energy intensive industries use relatively small amounts of intermediate materials, explaining the relatively small demand effect.

Changes in an industry’s total value added could be due to changes in the number of plants or in average output per plant. Oil shocks appear to primarily affect average production per plant. The demand and supply effects cause similar changes in value added per plant as in value per industry. In comparison, a price increase causes a small, though precisely estimated, decrease in entry, and has no effect on exit. Most of the effect of an oil shock is to decrease plants’ output levels.

2. Adam Saunders, MIT, Information technology and organizational capital
download paper

Erik Brynjolfsson, MIT, Lorin Hitt, University of Pennsylvania, and Adam Saunders, MIT

ABSTRACT:
Our hypothesis is that over the last five years, firms have been reaping the delayed benefits of large investments made in technology-enabled organizational capital. Organizational capital refers to the complementary investments, such as the creation of new business methods, that were made alongside technology investments in the late 1990s, but have taken much longer to bear fruit. We investigate this phenomenon by combining our own data with that of the Census Bureau in order to find new ways of measuring organizational capital and to determine how firms can best take advantage of technology. Our research may shed light on the nature of the recent productivity revival and clarify the factors that are most important to its future sustainability.

Using an estimating framework that relates the market value of a firm to the stocks of various assets (including property plant and equipment, computers, organizational assets and other balance sheet assets), we can estimate the implied market value of various types of assets. While we find that, for example, the market values one dollar of installed property plant and equipment at very close to one dollar, computer assets consistently show larger market valuations – to the extent of exceeding $10 per dollar of installed capital stock. It appears that this “excess valuation” of computers is concentrated in firms that simultaneously make substantial investments in organizational capital (as we measure it) along with their computer investments.

Our preliminary results using Census data are very encouraging and are in line with our earlier work. We demonstrate that human capital and decentralization are associated with higher market value when combined specifically with IT investments. We do not find that these higher market values accrue when these organizational characteristics are combined with other assets, such as property, plant and equipment. This provides strong evidence that these technologies and these organizational practices are economic complements, and this has important implications for data gathering, measurement, and management practice. We are working to extend these results to additional measures of organizational capital and develop different estimation approaches. Thus, we can further examine the productivity effects of these organizational assets and consider how this value accrues over time.

3. Yoonsoo Lee, Federal Reserve Bank of Cleveland, Geographic redistribution of U.S. manufacturing and the role of state development policy
download paper

Yoonsoo Lee, Federal Reserve Bank of Cleveland

ABSTRACT:
Competition among state and local governments to lure businesses has attracted considerable interest from economists, as well as legislators and policy makers. This paper quantifies the role of plant relocations in the geographic redistribution of manufacturing employment and examines the effectiveness of state development policy. Only a few studies have looked at how manufacturing firms geographically locate their production facilities and have used either small manufacturing samples or small geographic regions. This paper provides broader evidence of the impact of plant relocations using confidential establishment level data from the U.S. Census Longitudinal Research Database (LRD), covering the full population of manufacturing establishments in the United States over the period from 1972 to 1992. This paper finds a relatively small role for relocation in explaining the disparity of manufacturing employment growth rates across states. Moreover, it finds evidence of very weak effects of incentive programs on plant relocations.

Rocio Bonet, University of Pennsylvania

ABSTRACT:
Organizations play an important role on individual careers. Extant empirical evidence reveals that long-term attachments between workers and firms are common (Gibbons and Waldman, 1999) and that intra-organizational career mobility is a crucial phenomenon in most people’s work lives (Rosenbaum, 1979). Some studies have quantified that mobility between employers accounts only for one third of wage growth over the first ten years of experience (Topel and Ward, 1992). It is through wage growth and promotions within the organization that most people get ahead in their careers. What is it about firms that causes employees to advance much more quickly in some than in others is therefore an important question to ask.

Although the study of career attainment has been a topic of interest to several disciplines, a review of the literature reveals that it has been a field remarkably under explored. Furthermore, during the last decades American firms have experienced substantial transformation in their work organizational practices. The value of innovative work practices was discovered in the U.S. in the 80s and their use grew steadily over the 80s into the 90s (Appelbaum and Batt 1994; Shaw, 2004). Firms have moved toward the use of practices such as teamwork, problem-solving teams, information sharing, job rotation, cross training, TQM and incentive pay (Cappelli and Neumark, 2001). These practices are supposed to revolutionize the old internal labor market model. Therefore, looking at how innovative work practices have affected wage growth and promotions is an important question to understand the career dynamics of individuals in the current organizational context.

In this paper we use a U.S. matched employer-employee dataset, the National Employer Survey, to analyze organizational effects on individual wage growth and promotions. Our data come from establishment-level surveys conducted by the U.S. Bureau of the Census for the National Center on the Educational Quality of the Work Force. The data has the advantage of having a component being a matched employer-employee data. The data has also a longitudinal component on the establishments that allows us to partially address some causality issues.

We use two different measures of career advancement: wage growth and number of promotions. We use the nature of the matched employer-employee data to first investigate whether there is an employer effect on employee wage growth and promotions even after controlling for individual characteristics. In particular, we estimate two different models, one for wage growth and one for number of promotions, as a function of an establishment fixed effect and observed individual characteristics (such as human capital variables and demographics). We find that there is a significant establishment effect on wage growth and promotions even after controlling for observed employee characteristics. We then estimate the same models substituting the establishment fixed effects by innovative work practices and other firm characteristics. The first interesting result is that the effects differ for wage growth and promotions, so the old model, where wage growth used to be attached to changes in levels does not seem to hold any longer. We find that innovative work practices such as self-managed teams, job rotation and employer provided training seem to be positively related with wage growth and promotions. This suggests that innovative work practices are not dead-ends for employees but actually provide a path for advancement. Discussion and limitations of the analyses are presented in the paper.

2. Tetyana Shvydko, University of North Carolina at Chapel Hill, Labor market rigidities and the employment behavior of older workers
download paper

David Blau and Tetyana Shvydko, University of North Carolina at Chapel Hill

ABSTRACT:
An abrupt transition from full-time work to complete retirement is by far the most common pattern of employment behavior at older ages. However, if tastes for leisure increase gradually at older ages, then the optimal retirement plan is to gradually reduce hours of work. This observation suggests the existence of labor market rigidities that constrain the employment behavior of older workers.

A wide range of studies have explored the discontinuities in retirement behavior associated with the current structure of Social Security, Medicare, private pension and employer-provided health insurance plans. Not much research focused on the demand side of the labor market. Limited opportunities for part-time and flexible-hours, low wages and reduced benefits for available part-time work, and inadequate training and promotion opportunities for older workers both at their career employers and at potential new employers may arise from fixed costs of employment, requirements of team production, the existence of firm-specific human capital, costly monitoring of worker effort, and other features of the technology of production.

These possible sources of labor demand rigidity are caused by features of the technology of production, and in some cases would affect all workers, not just older workers. But if the hours-of-work preferences of older workers differ systematically from those of younger workers, then the existence of technology-induced rigidities will be manifested in the age structure of a firm’s work force: the more important are rigidities, the lower the share of older workers at a firm. Thus, while technology cannot be measured directly, with firm-level data it may be possible to detect evidence of technology-based rigidities if such rigidities are manifested in differences in the age structure of the work force across firms. Our analysis will be facilitated by the use of a unique matched employer-employee dataset. Specifically, we merge longitudinal data on individuals from the Survey of Income and Program Participation (SIPP) with data on their employers from the Longitudinal Employer-Household Dynamics (LEHD) files. Access to the confidential matched worker-firm data is available at the Triangle Census Research Data Center at Duke University.

We use a difference-in-difference approach to analysis: compare the behavior of older and younger workers in establishments with a larger share of older workers to the behavior of older and younger workers in establishments with a smaller share of older workers. Detailed data on workforce composition of establishments allows us to experiment with alternative definitions of large and small proportions of older workers employed by establishments. Taking the difference between the employment behavior of older and younger workers makes it possible to disentangle the effects of labor market rigidities that affect all workers from those that are specific to older workers. However, if there are factors other than labor market rigidities that cause the age composition of employment to affect the behavior of older workers differently than that of younger workers, this would threaten the validity of the difference-in-difference design. We control for pension and health insurance coverage and the worker’s wage rate, industry, occupation, and location to partially address this problem.

3. Richard V. Burkhauser, Cornell University and Shuaizhang Feng, Shanghai University of Finance and Economics, The Effects of Top-coding on Estimates of United States Wage Earnings and Income Inequality
download paper

Richard V. Burkhauser, Cornell University and Shuaizhang Feng, Shanghai University of Finance and Economics

ABSTRACT:
The levels of wage, earnings, and income inequality and their trends are of considerable interest to both policy makers and the general public. In the United States, the Current Population Survey (CPS) March Supplement is the primary data set used to measure these trends. However, researchers typically only have access to the public-use version of the CPS. Unlike the restricted-access data that is available to Census Bureau appointed researchers, income information in the public-use data are top-coded to protect confidentiality. This has lead researchers to either consistently top code the public-use data to control for this problem (e.g. Karoly and Burtless, 1995; Burkhauser et al., 2004) or to use a P90/P10 ratio because it is less susceptible to top coding issues (e.g. Autor, Katz and Kearney, 2005; Gottschalk and Danziger, 2005; Daly and Valletta, 2006).

However, Feng, Burkhauser and Butler (2005), using public-use CPS data, show that there are likely to be “top coding” problems in the restricted-access CPS data since over its history the CPS has increasingly captured the upper tail of the distribution of all sources of income. This, together with time-consistency problems of top coding in the public-use data, means that users of both the public-use and restricted-access CPS are likely to understate the level of wage earnings and income inequality in earlier years and overstate their growth over time. They provide a solution to this problem by modeling the wage earnings of full-time workers using the generalized beta distribution of the second kind (GB2), calculating Gini coefficients from the estimated parameters, and comparing them with past findings.

In this paper we extend the work of Feng, Burkhauser and Butler (2005) based on our access to the restricted-access CPS data. We first show that the use of unadjusted P90/P10 ratios based on the public-use CPS closely match P90/P10 values using the restricted-access CPS data before 1989 (because while there is censoring, virtually none of it occurs for those below the 90th percentile) but understate the restricted-access values thereafter because the percentage of censored income sources that fall below the 90th percentile of the income distribution rise substantially after 1989 due to the top coding of non-government income sources. However, we show that the problem can be solved by using cell means based on the restricted-access data. Because the Census Bureau does not provide cell means for all years in a consistent manner, we develop an alternative set of cell means for all CPS years and show that their use will result in P90/P10 ratios using the public use data that better match those based on the restricted-access data.

We then compare trends in our corrected public-use P90/P10 ratios to trends in our GB2 estimated Gini values (based on public-use data) for the wage earnings of all full-time workers, all earnings of full-time workers, and the household size-adjusted income of the entire population, and show that inequality trends based on these P90/P10 ratios do a poor job of matching trends in our GB2 Gini values.

4. Till M. von Wachter, Columbia University, Estimating the “true” cost of job loss: Evidence using matched data from California 1991-2000
download paper

Andrew K.G. Hildreth and Elizabeth Weber Handwerker, University of California-Berkeley, and Till M. von Wachter, Columbia University

ABSTRACT:
The paper examines problems in measuring and estimating the displacement of workers and the associated cost of job loss. There are broadly two approaches to estimating the cost of job loss for workers in the literature. One using the Displaced Worker Supplement (DWS); the other using state administrative records (chiefly the Unemployment Insurance base wage file). While are problems and benefits with both approaches, the estimates of the cost of job loss are markedly different. This paper uses a unique matched data set between the DWS and the Unemployment Insurance Base Wage (UI-BW) file for California. If the information content of the DWS is compared to the UI-BW, it appears that either the DWS under counts, or the UI-BW over counts the number displaced. Correcting for measurement error in the displacement indicator (that is correlated with workers age), the results indicate that the ‘true’ cost of job loss lies somewhere between -16 and -12 percent of their pre-displacement wage. Workers displaced because their plant closed do not suffer the same cost of job loss. The ‘true’ cost of job loss, for workers displaced from a plant closing, appears to be about -2.5 percent of their pre-displacement wage.
 

Yizhao Yang, University of Oregon

ABSTRACT:
Higher density, mixed land uses and housing, as well as better connected street network, are seen as the physical expression of some of the American urbanism principles: compact, diversity, and connectivitiy (Talen 2005). These physical environmental characteristics are now at the center of current planning debate regarding the desirable way of future urban growth. Proponents of the New Urbanism believe they should be part of a formula for building healthier and more sustainable communities and regions (Calthorpe 1993, Duany et al. 2000); proponents of Smart growth view them necessary in order to mitigate the development pressure from growth control (Danielson et al. 1999). Opponents, on the other hand, claim that such kind urbanism is inconsistent with Americans’ housing preference. So far academic debate and policy formulation have not benefited from systematic analyses of residents’ experience in such kind of physical environments.

In this paper I examine whether the aforementioned urbanism characteristics impose any systematic influence on residents’ satisfaction with their residential environments. Specifically, I study the extent to which housing consumers’ perception and rating of their residential neighborhood as a place to live varies according to the level of urbanism expressed in the physical environments.

I link household-level records in the Miami MSA from the 2002 American Housing Survey (AHS) to data from the 2000 decennial Census and Census Transportation Planning Package, as well as street pattern indicators generated using TIGER GIS files. I use various indicators for densities (housing, population, and employment densities), indexes of qualitative variation (IQV, for land use mix and housing mix), as well as street connectivity indicators, to gauge the physical expression of urbanism at multiple spatial scales including one’s immediate residential environment (as housing block), more general neighborhood (as census tract). I use mixed linear model that takes into account the hierarchical nesting nature of residential environments to examine the effects of urbanism characteristics.

Analysis results indicate that urbanism at different spatial scales exhibit different influence on people’s evaluation of their residential neighborhoods. Some urban characteristics are associated with greater neighborhood satisfaction while others are not. There also exist interactional effects between urbanism at different spatial scales. Findings hint to a need for more careful spatial analysis of people’s residential experience in the debate about neighborhood design.

Qingfang Wang, University of North Carolina at Charlotte

ABSTRACT:
A considerable amount of literature has addressed the issue of women frequently working in the semi- or low- skill level of the labor market sectors with unstable working conditions and insufficient pay. Immigrant women in particular face greater hardships in the labor market when compared with native women and immigrant men. The “spatial entrapment of women” hypothesis argues that women’s domestic responsibilities restrict their employment prospects and job-search areas, thereby spatially entrapping them in their neighborhoods of residence. Whereas most of the existing literature is based on the experiences of white women in western society (mostly Europe and North America), very little is known about the extent to which immigrant women face the same “spatial entrapment” and how space matters in their labor market process. This calls for an integrative study on the interaction between ethnicity, gender, and space in local labor market contexts.

Using a confidential dataset extracted from the Decennial Long Form Census Data 2000 and a multilevel regression modeling strategy, this research presents a case study of Chinese male and female immigrants in the San Francisco metropolitan area and examines how the geography of Chinese residences and workplaces influences their decisions to work in ethnic niches. The results show that in the San Francisco Bay area, living in Chinese residential concentrations and working in Chinese dominated work sites are strongly related to the probability of niche employment, although the direction and magnitude are different for men and women. The findings suggest that abundant ethnic resources in ethnic neighborhoods and enclaves can provide certain types of labor market opportunities; however, it also indicates the limitation of these resources in helping ethnic minority or immigrant workers, especially women, to move upward in the labor market hierarchy.

3. Jacob L. Vigdor, Duke University and NBER, When are ghettos bad? Lessons from immigrant segregation in the United States
download paper

David M. Cutler and Edward L. Glaeser, Harvard University and NEBR, and Jacob L. Vigdor, Duke University and NBER.

ABSTRACT:
Recent literature on the relationship between ethnic or racial segregation and outcomes has failed to produce a consensus view of the role of ghettos; some studies suggest that residence in an enclave is beneficial, some reach the opposite conclusion, and still others imply that any relationship is small. This paper presents new evidence on this relationship using data on
first-generation immigrants in the United States. Using average group characteristics as instruments for segregation, controlling for individual characteristics and both metropolitan area and country-of-origin fixed effects, we find evidence of a positive effect of segregation on outcomes, and of significant heterogeneity in this effect, with segregation more beneficial for groups with higher levels of human capital. A second analysis using an immigrant's actual neighborhood concentration, rather than the average figure represented by the typical segregation index, yields substantially different results. The benefits of segregation thus appear to accrue primarily to group members who do not themselves reside in the enclave community.

4. Stephen M. Schnebly, Arizona State University, The influence of community-oriented policing on crime reporting behavior
download paper

Stephen M. Schnebly, Arizona State University

ABSTRACT:
Community-oriented policing (COP) and crime reporting behavior are two topics that continue to garner considerable attention from criminal justice scholars and practitioners. Given the theoretical and practical connections shared by these research areas, the fact that each is a common subject of criminal justice research is of little surprise. What is surprising, however, is that these popular, substantively important, and related avenues of study have not yet explicitly intersected with one another. The present study addresses this gap in the extant literature by examining the influence of city-level police involvement in COP on multiple facets of crime reporting behavior, including whether a crime report was made, who made the crime report (the victim or a third party), and to whom was the crime report directed (the police or non-police officials). The study’s primary dataset, which was made available under special agreement with the U.S. Census Bureau, is the 1997-99 Area-Identified National Crime Victimization Survey (NCVS). In addition to the incident-level victimization data found in the public-use NCVS files, the Area-Identified NCVS includes codes that identify the precise geographic locations of respondents’ households. These geographic identifiers were used to link NCVS data on incidents of serious interpersonal violence with city-level COP data obtained from Law Enforcement Management and Administrative Statistics: Sample Survey of Law Enforcement Agencies (LEMAS: SSLEA). These city-level COP measures include the percentage of officially designated COP officers, the extent of efforts to train officers and citizens in COP, and the formation of problem solving partnerships between the police and the public. To ensure that any observed effects of COP on reporting behaviors were not confounded with other city-level characteristics, the study also incorporated annual homicide rates computed from Uniform Crime Reporting Program (UCR) data and socioeconomic and demographic data obtained from the 1990 and 2000 decennial censuses. Results generated from the study’s multinomial logistic regression models reveal that net of individual-, incident-, and city-level characteristics, COP does significantly influence some reporting behaviors. For instance, while the likelihood of third party police notification (relative to no report) is significantly greater in cities with a relatively large proportion of COP officers, victims residing in such cities are significantly less likely to report their victimization experiences to the police than they are to report to non-police officials. The findings also reveal that in cities where the training of current officers in COP is relatively extensive, victims are more likely to report to the police than they are to notify non-police officials or not report at all.