Recipient Organization
UNIVERSITY OF CALIFORNIA, BERKELEY
(N/A)
BERKELEY,CA 94720
Performing Department
(N/A)
Non Technical Summary
Research into agricultural productivity often abstracts from quality. In particular, when assessing the productivity of agricultural labor researchers often focus on a worker's speed rather than the value of their output.We explore four research questions: (1) Does Haste Make Waste? We use observational data and an instrumental variables research design to investigate the trade-off between worker speed and output quality. (2) We study the effect of paying direct quality incentives on a worker's speed and quality. (3) We estimate the costs of quality by measuring how quality affects the probability of a retailer refusing an order for quality concerns. (4) We extend our results to five hand harvested crops, exploring optimal contract structures for each.Our project builds on a unique two-way partnership between the co-PIs and a large strawberry grower-shipper. The partnership facilitates unprecedented access to data on worker speed, output quality, grower shipments and retailer refusals. In return, the researchers contribute to the grower's thinking about how to structure contracts to reward quality and maximize profits, which has led to a unique opportunity to evaluate novel contract structures. These extraordinary data will allow us to advance the research frontier on worker productivity by empirically assessing the economic mechanisms that drive quality and quantity. We will adapt existing models of productivity and test generated hypotheses using modern causal econometric methods. Our research is directly relevant to the Economics, Markets, and Trade (A1641) program areas in "farm labor and immigration and policy" and "agricultural production and resource use".
Animal Health Component
25%
Research Effort Categories
Basic
75%
Applied
25%
Developmental
(N/A)
Goals / Objectives
We have four objectives for this project:Objective 1: Does Haste Make Waste? Quantify the speed and quality trade-off using observational data.The aim of objective 1 is to quantify the speed and quality trade-off using two seasons of already collected data provided by the participating grower. We have completed the manuscript associated with objective 1.Objective 2: Direct Quality Incentives. Estimate speed and quality effects of paying workers for quality.The aim of objective 2 is to estimate the effects of quality incentive payments on both speed and quality. This objective requires real time monitoring of our quasi-experimental intervention consisting of paying a treatment group of harvesters different forms of a quality bonus over 3 growing seasons.Objective 3: Refusals. Quantify the value of quality in terms of the marginal effect of quality on the probability of shipment refusals.The aim of objective 3 is to quantify the short-term value of quality to growers. To do so we will link our data on quality with sales data provided by the retailer to estimate the effect of quality on the likelihood a shipment is rejected. We will then use the cost associated with each rejection to quantify the marginal cost of quality.Objective 4: Optimal Payment Structures. Use results from objectives 1- 3 to determine optimal payment schemes and pay rates.The aim of objective 4 is to use results from objectives 1 and 3 to estimate the short-run financial impacts of piece rate increases, and alternative payment structures (i.e. quality incentive payments) for employers in a variety of agricultural settings.?
Project Methods
Objective 1: Trade-off Identification using Observational Data - Simultaneous System of Equations ApproachTo obtain estimates of the trade-off between speed and quality, we propose using a simultaneous system of equations to model how two exogenous mechanisms - the speed of other workers (peer effects) and increases in the piece rate - simultaneously influence the speed and quality of a focal worker. We plan to model the effects of peer speed following the peer effects two- stage, fixed effects identification approach in Mas and Moretti (2009) and Hill and Burkhardt (2021). We plan to model the effects of two mid-season piece rate increases using a regression discontinuity in time (RDiT) approach - comparing speed and quality in the windows immediately before and after the increases. We detail these empirical approaches in our project narrative, but do not include the equations here for simplicity. The speed and quality trade-off can then be estimated as the ratio of coefficients from our simultaneous system of equations - the effect of peer speed on worker speed divided by the effect of peer speed on worker quality and the effect of the piece rate increases on worker speed divided by the effect of the piece rate increases on quality. Constructing trade-off estimates in this manner enables us to compare effects from these two mechanisms, and serves as a robustness check in that if workers are trading off between speed and quality, the magnitude of the trade-off should be consistent regardless of the mechanism.Objective 2: Direct Quality Incentives. Estimate speed and quality effects of paying workers for quality.To estimate the effects of our quasi-experimental quality incentive payments, we propose a differences-in-differences (D-i-D) research design. D-i-D designs compare differences in outcomes - say speed - between crews that are selected to receive the new quality incentive and control crews, before and after the new payment scheme is implemented. D-i-D estimates isolate the causal effect of interest, here the effect of a quality and/or speed-based compensation policy, on the outcomes of interest. The key (testable) assumption is that both treatment and control crews experience similar trends before treatment. As in objective 1, our outcomes of interest are speed and quality and our regression approach will include a time fixed effect to control for unobserved differences across days that might affect the outcome variable, for example weather, a worker-specific fixed effect term to absorb any time-invariant characteristics of workers that might influence how they respond to the new incentive payment, and a vector of additional covariates to control for other potentially confounding factors, including time-varying characteristics of fields, each worker's cumulative picking experience, and the (time varying) compositionof a worker's crew. As in objective 1, we will estimate the trade-off between speed and quality as the ratio of coefficients from the speed and quality regressions, i.e. the effect of the quality incentive payment on speed divided by the effect of the quality incentive payment on quality.Objective 3: Valuing QualityWe will estimate a flexible semiparametric dose-response model - this captures the intuition that the effect of an incremental defect when the overall quality of a shipment is high may be different than when quality is low. In other words, we do not expect the effect of a change in quality to be linear over the range of quality observed. We will consider multiple versions of quality for the explanatory variable, including specific quality defects (e.g. bruising, over-ripeness, and decay) and multiple measures of absolute and relative quality (e.g. minimum, maximum, and average quality for each order). We will estimate the model using a local constant least squares approach and compute partial effects of the individual quality components using the approach proposed by Henderson and Parmeter (2015). This will yield a dose-response function capturing the relationship between each quality metric and the likelihood a shipment is rejected. The relative magnitude of the coefficients indicates which quality metric is the strongest predictor of refusals. We include details on this approach in our project narrative, but omit the equations here for brevity.Objective 4: Optimal Payment Structures.We have developed a simple conceptual framework to model the expected profits for example shipments in our setting. Determining the optimal payment structure involves using estimates from objectives 1-3 as parameters in the expected profits equation. In particular, these estimates determine the marginal cost per tray (a function of the speed and quality payments), expected quality, and the probability of rejections. We will compute expected profits for each of our considered payment schemes using these parameter values. We will assess the robustness of our findings by considering a range of possible inputs for each parameter in the expected profits equation. This will shed light on differences in optimal payment schemes for different shipment sizes, prices, transportations costs, marginal costs, and impacts of quality on rejections. We will extrapolate these results to produce reports on optimal payment schemes for some of the major hand harvested crops produced in Colorado and California, including peaches, grapes, raspberries, apples, lettuce, and avocados. We will gather information on typical harvest wages and worker speed using the UC Davis cost and return studies for these agricultural commodities, supplementing these reports with conversations with growers, extension agents, and personnel at the High Planes Intermountain Centers for Agricultural Health and Safety (HICAHS). We will then develop bounds on potential changes in expected profits, worker speed, output quality, and worker incomes for each of the of compensation policies in our study, over a reasonable range of parameter values for harvest wages, the speed and quality trade-off, and the effect of quality on product rejections.?