Comments on “Customer-Employee Substitution:

Transcript:Comments on “Customer-Employee Substitution::
Comments on “Customer-Employee Substitution: Evidence from Gasoline Stations” Emek Basker, Lucia Foster and Shawn Klimek Susan Houseman, Upjohn Institute for Employment Research Presentation prepared for NBER-CRIW meetings, Cambridge, MA, July 18, 2017 Overview Provocative paper on important topic: Customers may substitute for workers in tasks that are manual, cognitively simple, but difficult to automate Can result in sizable worker displacement Estimated one-third fewer “attendants” in gas stations Customer-employee substitution can significantly inflate measured productivity levels/growth Back-of-the-envelop estimates: productivity growth inflated by 12.5% from 1977-1992 Divide comments into 3 areas: Empirical work on gasoline stations Broader measurement challenge of customer-employee substitution in retail Analogy to measuring labor productivity when tasks outsourced Current prevalence of gas station attendants in states with and without bans Source: BLS, Occupational Employment Statistics Program If gas station production function does not systematically differ across states, implies full service stations employ 24% (Oregon) to 42% (NJ) more workers. Effects of customer-employee substitution in gas stations Overall approach sensible given data limitations Endeavor to estimate effects on “attendant” employment, but lack occupation data Attempt to control for employment associated with ancillary services provided at gas station (repair, convenience store) Stations that are part of multi-unit or vertically integrated firm may have different employment composition—e.g., some administrative services provided at headquarters. Controls likely incomplete/quantity gas sold or # self-service pumps could be picking up effects on other services Regression estimates sensitive to inclusion of control variables Estimated effects on attendant employment would be greater today, with pay-at-pump Effects of customer-employee substitution in gas stations (cont.) More robustness checks on estimates of employment effects desirable Could exploit data from states with bans to estimate employment effects States with bans used in estimating back-of-envelop productivity effects Critical assumption that service provided by full and self-service gas stations, and so can be captured by gas sales, is debatable Full-service stations provide a bundled service Wash windshields, check oil Perform task of pumping gas Customer-employee substitution in retail Many rapidly developing cases in retail Scanner technology allows customers to check themselves out, replace cashiers On-line shopping, replace sales staff On-line banking, replace tellers Measuring effects on employment/worker displacement May not be easy because of data constraints Relatively straightforward conceptually Measuring effects on productivity—trickier, raises complicated conceptual questions How do we measure customer labor input? How should wait time, time driving to the retail outlet be factored in? Even if customers performing some tasks,