آیا زنان واکنش متفاوتی به انگیزه ها نشان می دهند؟ شواهدی از داده های تجربی و سوابق حقوق و دستمزد

We consider the response to incentives as an explanation for observed productivity differences within a firm that paid its workers piece rates. We provide a framework within which the observed productivity differences can be decomposed into two parts: One due to differences in ability and the other due to differences in the response to incentives. We apply this decomposition to male and female workers from a tree-planting firm in British Columbia, Canada. Our evidence (both experimental and nonexperimental) suggests that there was no difference in the reaction to incentives between male and female planters. The productivity differential male planters enjoyed arose because of differences in ability, interpreted as strength in our application.

Economists have shown great interest in measuring and explaining differences in the labour market performance between men and women. Traditionally, attention has focused on the earnings premium enjoyed by men and its relationship to productivity differences; see, for example, Gunderson (1989). Following Oaxaca (1973), the empirical implementation of human capital models permits the decomposition of the earnings premium in two parts: One explained by differences in economic characteristics and the other a residual. Empirical results indicate that the residual is important; see Blau et al. (1998). While these results are consistent with labour market discrimination, they may also reflect behavioural differences between groups; see Bowles et al. (2001). Men and women may differ with respect to motivation and aggressiveness, unobservable attributes that could be reflected in labour market outcomes; see Becker (1985) as well as Hakim (2000). In this paper, we develop a framework that allows us to test for differences in motivation, the response to incentives, between men and women. We show that productivity differentials can be decomposed into differences in ability and differences in the response to incentives. We apply this decomposition to experimental and nonexperimental data concerning men and women working in a tree-planting firm located in the province of British Columbia, Canada. The workers in this firm are paid using piece-rate incentive contracts: Payment is strictly proportional to individual worker productivity, the number of seedlings planted per day. Our data set contains information concerning daily worker productivity and the piece rate received per seedling planted as well as worker characteristics—gender and age. On average, male planters are more productive than female planters, by about 10%. Since this difference is conditional on planting conditions and the piece rate, we dismiss discrimination as an explanation and concentrate on inherent productivity differences between men and women. We exploit variation in the piece-rate to estimate the degree to which the observed productivity differential arises because of differences in the response to incentives. Payroll records containing variation in observed contracts coupled with information on actual productivity provide promising data from which to measure the response to incentives.1 Yet the variation in contracts observed within payroll records is typically not exogenous: The particulars of a contract are choice variables of firms. If elements (unobserved by the econometrician) affect the choice of contract as well as worker productivity, then the observed covariation between the contract and productivity will confound incentives with changes in unobservables: Reduced-form regression methods cannot identify the incentive effect.2 Tree-planting firms choose the piece-rate as a function of planting conditions that are unobserved by the econometrician, yet affect worker productivity. Blocks of land to be planted are inspected in advance by the firm. Generally, the steeper is the terrain, or the greater is underbrush or rock on the ground, the more difficult it is to plant and the lower will be worker productivity and thus earnings. Since the contract with the workers contains only one instrument (the piece rate) it must perform multiple tasks—motivating workers and inducing workers to plant under different conditions. Consequently, as planting conditions become more difficult, the firm raises the piece rate to meet the labour-supply constraint. In fact, a simple regression of worker productivity on the piece rate yields a negative coefficient, reflecting the endogenous setting of piece rates. Our first approach to solving this endogeneity problem is structural. We use economic theory as a lens through which to interpret worker and firm decisions concerning effort choices and piece rates as functions of unobservable planting conditions. Our model is based on extensive interviews with firm managers during which the institutional setting was revealed and, most importantly, the manner in which piece rates are determined. Incorporating the derived decision rules directly into our estimation procedure allows us to measure worker reactions to incentives, controlling for unobservables. Our structural model adds to the recent literature concerned with applying structural econometric methods to data from firm records in order to investigate incentive issues; see Ferrall and Shearer (1999), Paarsch and Shearer, 1999, Paarsch and Shearer, 2000 and Paarsch and Shearer, 2005, Haley (2003), and Shearer (2004) as well as Copeland and Monet (2002). We extend previous models to allow for heterogeneous responses to incentives and concentrate on differences between men and women. Estimating structural econometric models that explain earnings differences between men and women was first proposed and implemented by Bowlus (1997) and later extended to black–white differences by Bowlus and Eckstein (2002). Behaviouralists describe incentive-enhancing preferences as those in which a worker's effort level is an increasing function of his reward; see Bowles et al. (2001). We focus on differences in parameters that determine a worker's cost of effort. In our model, this can happen two ways—either from differences in a worker's ability or from differences in a worker's reaction to piece rates. Differences in ability affect the effort level, but not the percentage change in effort with respect to the piece rate, the elasticity. Identifying differences in the response to incentives separately from differences in ability through structural estimation depends on the way these parameters affect the distribution of observed productivity. Our model admits identification because these two factors affect the variance of productivity differently. Individuals who react a great deal to incentives take advantage of favourable economic conditions to increase productivity: Ability affects average productivity, while reaction affects average productivity as well as the variance of productivity. We show that our framework allows us to rank workers in terms of their responses to incentives and to test whether these responses are homogeneous. Incorporating individual characteristics into the model allows us to test whether women react differently from men. Some have criticized the structural approach, highlighting the sensitivity of results to assumptions concerning the primitives of the structural model—the form of the utility and the production functions as well as distributions of unobservables. In our second approach, one involving a field experiment, we seek to measure the robustness of our results to departures from these assumptions. We exploit exogenous variation in piece-rate contracts provided by a small-scale field experiment to measure directly worker responses to incentives. During the field experiment, blocks of homogeneous land to be planted were divided into two sub-blocks—control and treatment. On the control blocks, workers received the piece-rates that the firm would normally pay on those blocks, as determined by planting conditions. On the treatment blocks, workers received a piece rate that was five cents higher than the control rate; this represented an increase of between 21% and 28% over the control rate. The exogenous variation in the piece rate during the experiment provides a direct way to measure worker response, free of the assumptions of the structural model. Moreover, interacting the piece rate with a dummy variable indicating planter gender provides a simple test of whether men and women react differently to incentives. Our experimental results are very similar to the structural results: Women respond slightly more to incentives than men, but the difference is not statistically significant. The similarity between the structural and the experimental results provides some assurance that structural results are not a pure consequence of functional-form assumptions. We also provide evidence that the structural model has some ability to predict the individual-level responses to incentives that are observed in the experiment. Our paper is organized in seven more sections. In Section 2, we provide some institutional details concerning the tree-planting industry in British Columbia, while in Section 3 we describe our payroll data. In Section 4, we develop a model of worker effort decisions and present our productivity decomposition, while in Section 5, we present our structural results. In 6 and 7, we present the experimental data and results, and in Section 8 we discuss our results and conclude. In an appendix to the paper, we present the proofs of a lemma and two theorems too cumbersome to be included in the body of the paper.

Workers differ in productive characteristics, both observed and unobserved. Controlling for these differences is important when interpreting the causes of observed differences in economic outcomes as well as potential policies to overcome those differences. We have exploited variation in incentive contracts to identify differences in worker motivation (or responses to incentives) as a possible explanation for observed productivity differentials between men and women. Our results suggest that, while workers do respond differently to incentives, no systematic difference between men and women exists in this regard; the observed productivity differentials are attributed to differences in ability, strength in our application. Our statistical conclusions are based on two separate and distinct data sets and two different econometric strategies. Using unrestricted analysis-of-variance, we analysed data from field experiments under which exogenous variation in the observed contracts is guaranteed. Using structural econometric methods, we analysed data from payroll records, controlling for endogenous variation in nonexperimental contracts. Both approaches yielded similar results. The experimental results suggest that functional-form assumptions are not driving the conclusions of the structural estimation. Note, too, that the results appear robust across time. When we re-estimated the structural model using data from payroll records gathered nearly a decade after the original sample, we found similar results. This evidence suggests that the distribution of worker types has been relatively stable over the last decade. Our results suggest that the observed differences in productivity between men and women are solely attributable to differences in physical strength in our application. In environments where physical strength is unimportant, our results suggest that productivity differentials should disappear, consistent with the findings of Gneezy et al. (2003). A complete generalization of our results would require identification of the population distribution of elasticities. At the least, this would require estimating a model that incorporated both productivity decisions and the decision to participate in the firm. The development and identification of such models remains an important area for future research based on firm-level payroll records. Finally, our results provide an interesting contrast to historical work in the sociology literature. Some sociologists have argued that men and women react differently to incentives because of differences in their family environments. Shimmin (1962) noted that young female workers in England during the 1950s and 1960s often had their earnings expropriated by their families in return for a fixed allowance. She argued that these social conventions, which did not apply to men, dampened female responses to incentive payments; see Millward (1968) for some evidence of these effects. The contrast with our results undoubtedly reflects the changing circumstances under which women participate in the labour market today.