Labor Economics
Applied Econometrics
Personnel Economics

Labor Economics
Applied Econometrics
Personnel Economics
Microeconomics

October 1998 (Oral) Industrial Organization, Labor Economics (with distinction)
May 1997 (Written) Microeconomic and Macroeconomic Theory

Empirical Essays in Personnel Economics

Professor Michael Boozer
Professor T. Paul Schultz
Professor Ann Huff Stevens

Summer 2002

M. Phil., Economics, Yale University, 1999
M.A., Economics, Yale University, 1998
A.B., Magna Cum Laude, Princeton University, 1996

Recipient of the Raymond Powell Prize for Outstanding Teaching by a Graduate Student in
    the Yale Economics Department, 1999-2000
Yale University Graduate Fellowship, 1996-2000
Phi Beta Kappa, Princeton University, 1996

Teaching Assistant, Introductory Macroeconomics, Yale University, Spring 2001
Teaching Assistant, Econometrics, Yale University, Spring 2000, Spring 1999
Teaching Assistant, Introductory Microeconomics, Yale University, Fall 1999, Fall 1998

Research Assistant to Professor Michael Boozer, Yale University, 1998
Performed empirical work to analyze the impact of state-level school finance reforms on the distributions of student-teacher ratios.

"The Impact of a Monitoring Technology on Worker Incentives and the Coordination of Firm Activity: Evidence from the Trucking Industry," manuscript, Yale University, 2001. (job-market paper)

"Inside the 'Black Box' of Project STAR: Estimation of Peer Effects Using Experimental Data" (with Michael Boozer), Yale University Economic Growth Center Discussion Paper No. 832, 2001.

"Nonseparabilities, Endogenous Preferences, and Measurement Error in the Analysis of Intertemporal Labor Supply: Testing and Structural Analysis Via Correlated Random Effects" (with Michael Boozer), work in progress.

While the theoretical underpinnings of personnel economics and agency theory have been richly developed, rigorous empirical testing and evaluation of these models has lagged far behind. The first chapter of my dissertation contributes to the empirical evidence in this field by exploiting the introduction of a new technology that directly monitors worker behavior. Using a combination of unique data sets, I utilize both the time series and cross-sectional variation in the data to estimate the causal impact of the monitoring technology on several dimensions of worker and firm behavior. First, I provide direct tests of the value to firms of this new technology, and disentangle the component attributable to improved employee incentives from the element that enhances the coordination and resource allocation decisions in firm activity. Second, I examine whether observed compensation contracts vary in ways predicted by the theory, and I estimate how strongly agents react to changes in the monitoring environment.

The forum used to analyze these issues is the trucking industry, where the introduction of on-board computers (OBCs) in the late 1980’s had a tremendous economic impact. OBCs, which are small computers installed on individual trucks, are of two forms. The slightly older technology of trip recorders provides firms with information that can be used to improve incentives. The more modern and advanced Electronic Vehicle Management Systems (EVMS) allow for a real-time exchange of information between drivers and firms that enhances both incentives and coordination decisions. The previous literature has attempted to disentangle these incentive and coordination effects by comparing use rates of trip recorders and EVMS across different sectors of the trucking industry. Using a very detailed survey of truck physical and operating characteristics from the Census of Transportation, I improve on the previous literature by more carefully considering the variation in OBC adoption. I argue that in order to separate the incentive and coordination effects, one must take account of the differential replacement rate of trucks by sector. New trucks come bundled with the current form of OBCs, leading to a secular trend of EVMS use across model years. To factor out this secular increase in EVMS adoption, I show that we need to rely on the within model year variation in OBC use, as the between model year variation is corrupted by truck age and replacement. After this adjustment is made, the incentive effect of OBCs remains, while the coordination effect becomes very difficult to detect.

While the approach described in the preceding paragraph is useful for establishing evidence "consistent with" certain theoretical propositions, the model that I establish in the paper suggests a more direct means of separating the incentive and coordination effects of OBCs. The coordination effect (which includes improved quality of customer service via shipment tracking capabilities) should influence revenues as firms with EVMS become more efficient in allocating trucks and drivers to hauls. The incentive effect, on the other hand, should influence the cost side of firms’ accounts, since improved driver effort effectively reduces the cost per unit of labor. Using cell-level data created from the merger of firm-level financial data with the OBC data, I use the differential effects of trip recorder and EVMS adoption on firm revenues and costs to separately identify the monetary value of the incentive and coordination improvements of OBCs.

The second half of this paper tests several implications of principal-agent theory. Of particular relevance to a truck driver’s compensation contract is the Holmstrom-Milgrom (1991) multitask principal-agent model. In the trucking context, one of the driver’s tasks is easily observed without an OBC, but the effort directed towards his other task becomes much more precisely measured upon adoption of an OBC. The theoretical model presented yields the standard result that as the variance in observed output not attributable to agent effort falls, then the incentive intensity should increase. Since OBCs allow for a reduction in this variance, sectors of the industry with high OBC use are predicted to use "higher-powered" contracts that tie compensation more closely to driver performance. Using cell-level data constructed from combining the OBC data with a data set rich in firm-level information on driver contracts, evidence is presented which confirms this prediction. Moreover, the incentive intensities on both tasks are larger when OBC adoption is greater, indicating that sectors with low OBC adoption mute the incentive intensity on the more observable task in order to prevent the agent from neglecting the less observable task.

Given the more precise measurement of driver effort under the adoption of an OBC, and the attendant changes in the observed (and unobserved) components of the compensation contract, I then test for the impact on driver behavior. The theoretical model predicts that monitoring of drivers should lead them to adjust the mean and variance of their driving speed in a way that is more aligned with their employers’ objectives. This suggests two avenues to detect changes in driver behavior: first, an OBC should increase the life expectancy of a truck, and second, an OBC should increase a truck’s fuel efficiency. By constructing a synthetic panel using three waves of the Census data, I carefully consider the timing and endogeneity issues involved in uncovering the causal relationship between the dependent and explanatory variables. The empirical work presented provides direct evidence in support of the above propositions regarding driver behavior.

My second chapter, written jointly with Michael Boozer, is entitled "Inside the ‘Black Box’ of Project STAR: Estimation of Peer Effects Using Experimental Data." The credible identification of endogenous peer group effects -- i.e., social multiplier or feedback effects -- has long eluded social scientists. We argue that such effects are most credibly identified by a randomly assigned social program that operates at differing intensities within and between peer groups. The data we use are from Project STAR, a class size reduction experiment conducted in Tennessee elementary schools. In this experiment, students were randomly assigned to either a small class or a regular-sized class in Kindergarten. As the experiment progressed through time, classes became comprised of varying fractions of students who had previously been exposed to the small class treatment, creating class groupings of varying experimentally induced quality. We use this variation in class group quality to estimate the spillover effect. We find that when allowance is made for this ‘feedback’ effect of prior exposure to the small class treatment, the peer effects account for much of the total experimental effects in later grades, and the direct class size effects are rendered substantially smaller.