Customer‐employee substitution: Evidence from gasoline stations

• Published date: 01 December 2017
• Author: Lucia Foster,Shawn Klimek,Emek Basker
• Date: 01 December 2017
• DOI: http://doi.org/10.1111/jems.12215

Received: 23 December 2015 Revised: 26 May 2017 Accepted: 19 June 2017

DOI: 10.1111/jems.12215

ORIGINAL ARTICLE

Customer-employee substitution: Evidence from gasoline stations

Emek Basker Lucia Foster Shawn Klimek

U.S. Census Bureau, Washington,DC, USA

(Email: emek.m.basker@census.gov;

lucia.s.foster@census.gov;

shawn.d.klimek@census.gov)

Commentswelcome. Any opinions and con-

clusionsexpressed herein are those of the

authors and do notnecessar ilyrepresent the

viewsof the U.S. Census Bureau. All results

havebeen reviewed to ensure that no confi-

dentialinformation is disclosed. A previous

versionof this paper circulated with the title

“Customer-LaborSubstition.” We thank

AndreaShepard, Ivan Png, and David Reit-

manfor the Lundberg data; John Nogle for

helpwit h historic Census of RetailTrade data;

CYChoi for help with the ACCRA data; and

Chris Adams,Phillip Antopitov, Saku Aura,

DavidAutor, Randy Becker, Jim Bessen, Roger

Betancourt, AndrewFoote, Anne Hall, John

Haltiwanger,Russell Hillberry, Henry Hyatt,

RonJarmin, Richard Kneller, Mark Lewis,

Florin Maican, Kristin McCue, Peter Mueser,

MatildaOr th, Ivan Png, Henry Schneider,

JimSpletzer, Patrick Warren, seminar partici-

pantsat IFN–Stockholm, Bureau of Economic

Analysis,George Washington University,Uni-

versityof Michigan, George Mason University,

FederalTrade Commission, the 2015 IIOC

(Boston),the 2015 BLS-Census Workshop,

andt he 2015 ComparativeAnalysis of Enter-

prise Data (Istanbul) forhelpful comments and

conversations.

Abstract

We document the adoption of self-service pumps in U.S. gasoline stations from 1977

to 1992. Using establishment-level data from the Census of Retail Trade over this

period, we show that self-service stations employ approximately one quarter fewer

attendants per pump, all else equal. The work done by these attendants has shifted to

customers, biasing upward conventional measures of productivity growth.

1INTRODUCTION

Increasingly in the service sectors, customers are performing tasks previously done by paid employees: they withdraw, deposit,

and transfer funds using automated teller machines (ATMs); book travel online without the aid of a travel agent and check in

using automated kiosks at airports; check out at supermarket “self-checkout” lines; use tablets to order and payat restaurants; and

access customer-service functions online or through an interactive voice response telephone call.1In these and other settings,

customer–employee substitution is a solution to the problem of outsourcing tasks that are resistant to computerization. These

manual tasks have proven challenging to automate and computerize because they require “situational adaptability, visual and

language recognition, and in-person interactions” and “must be produced and performed largely on-site or in person” (Autor,

2015, p. 12).2Although a person is needed to perform the task, that person need not be an employee. Indeed, the one person

who is guaranteed to be on site at the time of the transaction and to understand the customer’s wishes and needs is the customer

herself.

Published 2017. This article is a U.S. Government work and is in the public domain in the U.S.A.

J Econ Manage Strat. 2017;26:876–896. wileyonlinelibrary.com/journal/jems 876

BASKER ET AL.877

From the perspective of productivity measurement, customer–employee substitution poses a unique quandary. Whereas true

automation reduces labor input, customer–employee substitution merely substitutes one type of labor, which is observed, with

another type of labor, which is unobserved.

In this paper, we use a case-study approach to study the transformation of gasoline stations and the effects of integrating

customers into the production function. To this end, using station-level data from the Census of Retail Trade (CRT) over the

period from 1977 to 1992, we estimate a production function taking into account the specific features of that production: namely,

that a car, a gasoline pump, and a worker (who may be a paid employee or the unpaid driver of the car) must all be present at

the same place for a fixed number of minutes. We estimate two key parameters of this production function: the productivity of

attendants at full-service pumps and the productivity of attendants at self-service pumps. The ratio of these two allows us to

infer the labor input due to customers, which we put at approximately one quarter of the labor input in self-service stations. We

then estimate both physical and revenue measures of gas-station labor productivity, and demonstrate that ignoring the role that

customers play in the production function leads to upward bias in measured productivity.

The self-service format in gasoline stations has been traced to a Urich Oil station in Los Angeles in 1947 (Reid, 1984). The

same source estimates that only 1% of gas stations offered self-service pumps in 1969. The Census Bureau started collecting

information on self-service options at gas stations in 1972. (Establishment-level data are available starting in 1977.) Table 1

provides summary statistics from published CRT tables for 1972, 1977, 1982, 1987, 1992, and 1997. Over this period, the total

number of U.S. gasoline station declined, and the share of gas stations providing some self-service sales of gasoline increased

from 8% to 80%.3

The paper contributes to a broader discussion on productivity measurement, particularly in retail and services. Retail pro-

ductivity is hard to measure because output is multidimensional (see Ratchford, 2016, for a general discussion); in the language

of Betancourt (2005), retailers offer not only explicit goods for sale but also “implicit distribution services,” which are funda-

mentally unmeasurable (p. 6). In addition, unlike in the manufacturing sector, capital is typically measured as square footage,or

not at all; infrastructure and other capital investments are rarely, if ever, included in retail-productivity calculations. Customer–

employee substitution adds another dimension to the measurement problem by introducing a nonmeasured component. The

difficulty is not simply that customers participate in the production of final goods but that the nature of their participation has

changed over time. If we think of the customer’s work as another input into the production function, customer–employee sub-

stitution means that we are completely missing information on one input. If we think of the customer’s work as a component in

total labor, then customer–employee substitution means that we are mismeasuring the labor input. In either case, the aggregate

input measure is biased downward and this bias increases as more stations adopt self-service technologies. Thus, true produc-

tivity may increase, particularly due to a reduction in idle time, but measured productivity increases disproportionately more.

Moreover, the (multidimensional) output itself has changed as one of the bundled services (an attendant delivering gasoline into

the customer’s car) has disappeared.

We describe the data sets used in the analysis in Section 2. Section 3 provides background on the adoption of self-service and

describes reduced-form regression results correlating adoption with observed factors. Section 4 presents a theoretical framework

and estimates of the labor-saving impact of self-service stations. Section 5 demonstrates the bias in conventional productivity

measures that do not account for customer–employee substitution and for the other major trend in gas stations over this time

period, which is the bundling of convenience stores into gas stations. Section 6 concludes.

TABLE 1 Trends in the gas-station sector, 1972–1997

1972 1977 1982 1987 1992 1997

Gas stations (SIC 554) 183,385 146,523 116,188 114,748 105,334 98,846

Convenience stores selling gas (SIC 541130) n/a 5,138 13,390 20,293 23,065 28,043

All locations selling gas n/a 151,661 129,578 135,041 128,339 126,889

Employment at gas stations (SIC 554) 747,668 672,673 603,886 701,690 675,080 741,040

Employment at convenience stores selling gas (SIC 541130) n/a 48,627 94,670 148,499 142,183 181,022

Total employment at locations selling gas n/a 721,300 698,556 850,189 817,263 922,062

Average station employment(SIC 554) 4.1 4.6 5.2 6.2 6.4 7.5

Average station employment(SIC 541130) n/a 9.5 7.1 7.3 6.2 6.5

Self-service (% of stations)a8.2 38.3 58.7 70.5 79.9 n/a

Source: Authors’ calculations from published and unpublished CRT data.

Note: Figures refer to stations with payroll.

aSIC 554 only.