Economists and Field Research: "You Can Observe a Lot Just by Watching"
Susan Helper 1 Forthcoming, American Economic Review, May 2000 Modern economics began with Adam Smith's visit to a pin factory, which helped him
explain how the division of labor worked (1985, p. 6). However, not many economists today do
much fieldwork, which involves interviews with economic actors and visits to the places they
live and work. To help economists get out more, in 1994 the Sloan Foundation funded the
National Bureau of Economic Research to promote field research among economists, via plant
tours, conferences, and comissioned papers. I start by discussing how fieldwork can improve economic research, drawing largely on
interviews with participants in the NBER/Sloan program. I then describe techniques that can
improve economists' field research. (For an expanded version of this paper, see
www.nber.org/sloan). I.Fieldwork complements other methods Economists today typically do research using econometrics and mathematical modelling.
These techniques have many strengths, but share the weakness of distance from individual
economic actors. In contrast, field research allows direct contact with them, yielding several
advantages: Researchers can ask people directly about their objectives and constraints. It is not always
easy to figure out someone's incentives or strategies by looking only at outcomes. For example,
Jim Rebitzer wondered why many professionals complain about long hours, yet few firms offer
the option of short hours. In talking with lawyers, he learned that partners found it difficult to
decide whom to promote in order to maximize their incomes. "One comment that stuck with me
was a partner saying about an associate, 'She does really good work, but I wonder, does she like
money enough?' That is, he wanted to know, will she work really hard? " Comments by this
partner and others implied that they used work hours as a proxy for this propensity to work. This
insight led Renee Landers, Rebitzer, and Lowell Taylor (1996) to build a model and collect
survey data which suggested that these observability constraints led to incentives to work
inefficiently long hours. Fieldwork allows exploration of areas with little pre-existing data or theory. I started my
dissertation research thinking I would look at automakers' make/buy decisions. But when I
started interviewing and reading trade journals, I realized that important changes--not reflected
in the existing literature--were occurring on the 'buy' side. US automakers were moving from
adversarial deals (in which they "would steal a dime from a starving grandmother", one supplier
said) to "voice" relationships in which they worked with suppliers to improve performance. My
qualitative study (1991) suggested that information exchange and commitment were important
determinants of supplier performance, leading me to collect survey data on these factors. One
finding was that voice relationships were associated with more cost reduction, but only if
complementary policies were adopted (Helper, 1999). Fieldwork suggested to Zucker, Darby, and Brewer (1998) that number of gene-sequence
discoveries was a good proxy for intellectual capital in biotechnology; their regression results
were consistent with intellectual capital being the main determinant of the location and growth
of biotechnology firms. Fieldwork facilitates use of the right data. Ichniowski, Shaw, and Prennushi (1997) used
interviews to determine that steel finishing lines had homogeneous technology and that there
were enough such lines to allow econometric investigation of the impacts of innovative human
resource policies unconfounded by technology or industry differences. They then visited 45
plants to collect production data (interviewing managers to insure compatible measures across
plants) and observe what human-resource practices were in place. Fieldwork provides vivid images that promote intuition. Edward Lazear said of his work
on the change from time-rates to piece-rates at Safelite Autoglass (1996), "It's one of my most-cited papers--I think it's because everyone can imagine those guys working harder to install
windshields once they're on piece-rates, and it's an image they remember a lot more than the
regression coefficients." Because of fears about the reliability of field methods (discussed below) some
economists get ideas from the field, but do not discuss their fieldwork in their published articles.
But this tactic causes us to lose the vividness that is a principal benefit of fieldwork.
Understanding the setting can help explain differences in findings between cases, by making
clear the mechanism by which variables are linked. For example, while Lazear found that a
move to piece-rates increased profits at the auto-glass installer, Richard Freeman and Morris
Kleiner (1998) found that a change away from piece-rates increased profits at a shoe
manufacturer. Understanding the production process at the two firms is key to making sense of
these results: While both papers found that productivity was higher under piece rates, time-rates
at the shoe firm facilitated the introduction of a new production process that brought reduced
inventories and faster new-product introduction. Many of these insights can be translated into the language of econometrics or theory. It is
possible that economists using only these methods could have generated these insights, but in
fact, they didn't. Fieldwork offers a new source of inspiration, one that is complementary to
more conventional methods. "It's important to go out and discover the facts for yourself," said
Ronald Coase, who developed his ideas about the "nature of the firm" (1937) during a year of
visits to firms throughout the US. II. Many criticisms of fieldwork can be answered with improved methods Many economists remain skeptical of qualitative research, fearing that it is not objective,
replicable, or generalizable. "We don't have standards for what good fieldwork is," said Eli
Berman. "In econometrics, we know to look for things like identification and specification
issues, but what are the analogues in field work? How is it different from journalism?" Better techniques can alleviate these problems. "Economists think that while econometrics
requires years of training, field research is easy, " said Rebecca Henderson. "But we need to pay
just as much attention to things like careful research design and sample selection as we do in
quantitative research." Below are tips from economists and others experienced in field methods.
(See Robert Yin (1989), Robert Thomas (1994) and www.nber.org/sloan for additional
suggestions. David Lodge (1990) presents information in novel form.) Objectivity. A way for economists to avoid confirmatory bias is to test hypotheses coming
from several competing theories. (Economists often go into the field with hypotheses to test. In
contrast, disciplines like anthropology emphasize understanding the world as their informants
do.) One source of alternative theories is to let respondents talk, even when you think they are
getting off the subject. You will probably learn something you never thought to ask about and
besides, getting to tell their story is part of their payoff for talking with you (Michael Piore,
1979). It is important to do more than look to confirm or disconfirm pre-conceived hypotheses.
Claudia Goldin said about a 1995 visit to Joseph Pollak, an auto-parts manufacturer in
Stoughton, MA: "I didn't have any particular expectations going into the plant, but I remember
vividly looking down from a mezzanine from which you could see the whole shop floor. As I
looked down, I realized I was observing--in one moment--the transition from 19th century
technology to 20th century technology. I could see the relative increase in the demand for skill
just scanning across the room. The 'continuous process' machinery required lots of skilled labor
to set up the machines and mechanics to maintain them; there were few operators...The
old-fashioned areas [making similar parts], however, were filled with semi-skilled workers and
almost no skilled workers. The scene sparked my imagination and I wrote two papers (Goldin
and Lawrence Katz, 1996, 1998). For years I had been reading the history of technology, but it
wasn't until I went to Pollak that I made the connection that adoption of continuous-process
technology was complementary to skill." Another way to enhance objectivity is to "triangulate", asking questions so that answers can
be checked against information from other interviewees, company documents, trade journals,
etc. Site visits also provide a way to get a perspective other than the one presented by
interviewees. A dramatic example of this came at our 1995 visit to LTV Steel in Cleveland,
when workers stopped us to explain their anger at management's participation in a non-union
joint venture in Alabama, which they thought violated the spirit of union-management
partnership. Managers hadn't mentioned this dispute to us in their presentation on determinants
of plant performance, even though it had caused the union to pull out of (productivity-enhancing) employee-involvement programs. This incident also illustrates the value of interviewing people in a variety of positions.
High-level people will give you an overview of the firm's intended strategy; low-level people offer
detailed examples of the incentives and constraints they actually face. Letting respondents tell their story does not mean taking everything they say at face
value--you should be as skeptical of their statements (and as appreciative of them) as of any
other data. Phrase questions concretely, and ask for examples. Until I asked for examples of
trustworthy behavior in customer-supplier relations, I didn't realize that the definition affected
the measured relationship between trust and performance. For example, one manager said, "I
have a very trusting relationship with the plant I supply. If they find defects, they'll call us up
and we can fix it, without anything ever showing up in the records." Another manager gave a
quite different description of a trustworthy customer: "They are incredibly strict on
quality--they'll send back a whole lot if even one part is defective. But they're always there to
teach us, to make us better." The first manager felt little incentive to improve from his
"trustworthy" customer; the second one felt a strong incentive. Replicability. Often interviewees won't talk freely unless they are promised confidentiality,
making it difficult for other researchers to replicate by interviewing the same people. However,
there are other ways to enhance replicability. Rebecca Henderson and Iain Cockburn (1994)
coded interview transcripts to create a dummy variable measuring whether pharmaceutical firms
had "pro-publication" policies for their scientists. (This variable was a significant determinant of
firm's research productivity. It is hard to imagine how else they could have obtained this
information, other than by asking the actors directly about their incentives.) The more clearly
described the coding, the more other researchers will be able to replicate it elsewhere. Care in writing up results can increase replicability and readers' confidence in qualitative
findings. For example, some field-research papers present arguments about what drives their
regression results, and then say "interviews confirmed these findings", without providing
information about how interviewees were selected, or enough information about what they said
for readers to judge for themselves. This is like asking readers to believe a summary of
econometric results without tables of regression coefficients. See David Levine (1993) on
describing selection of interviewees, Stephen Barley (1986) on using quotes as evidence, and
Kathleen Eisenhardt (1989) on using 'textual tables' to present this information concisely. Some changes at journals could alleviate the space problems that cause economists to
truncate description of their fieldwork. Authors could use journal web sites to post information
about field findings that didn't fit in the published article, and more journals could accept stand-alone qualitative analyses (the Journal of Financial Economics, Journal of Industrial
Economics, and Industrial Relations already do). Generalizability. How can one be sure that one's conclusions go beyond the few firms one
has visited? Careful selection of cases according to theoretical principles, and use of cross-case
analysis can alleviate this problem (Eisenhardt, 1989). If one is building theory, in-depth
understanding of a few cases may be appropriate. If one wants to use observations from plants as
data, visiting a large number of them is quite convincing (Ichniowski, et.al, 1997). As more
cases are done on a topic, the more we will understand what is specific to each setting, as the
piece-rate example shows. That is, the solution to the generalizability problem is to do more field
research, not less! No methodology is perfect. "Regressions also have serious problems of
generalizability (they predict poorly out of sample), subjectivity (researchers may stop
specification searches when their favorite t statistic rises over 2), and measurement error (critical
concepts like 'income' and 'capital' are very poorly measured). This is why research is hard--and
why we should believe only findings obtained with multiple methodologies," said Levine. In summary, good field research should be like good journalism in containing accurate,
vivid examples. It should go beyond journalism in explaining potential biases in selection of
cases, describing the construction of concepts used, and having a theoretical starting point and/or
outcome. In conclusion, field research can make us better economists--whatever our current
technique--by increasing our understanding of the objectives, constraints, and incentives that
economic actors face. REFERENCES Berra, Yogi. "Famous Yogi Berra Quotes." http://www.yogiberraclassic.org/quotes.htm, nd. Barley, Stephen. 'Technology as an Occasion for Structuring: Evidence from Observations of
CT Scanners. " Administrative Science Quarterly, 1986, 31, pp. 78-108. Coase, Ronald. "The Nature of the Firm." Economica, 1937, 4 pp. 386-405. Eisenhardt, Kathy. "Building Theories from Case Study Research." Academy of Management
Review, 1989, 14(4) pp. 532-550. Freeman, Richard and Kleiner, Morris. "The Last American Shoe Manufacturers." NBER
Working Paper 6750, 1998. Goldin, Claudia and Katz, Lawrence. "Technology, Skill, and the Wage Structure." American
Economic Review, May 1996, 96(2) pp. 252-257. _______. 'The Origins of Technology-Skill Complementarity." Quarterly Journal of Economics,
August 1998, 113 pp. 693-732. Susan Helper. "Strategy and Irreversibility in Supplier Relations." Business History Review,
65(4), Winter 1991, pp. 781-824. _______."Complementarity and Cost Reduction." NBER Working Paper 6033 (revised), 1999. Henderson, Rebecca, and Cockburn, Ian. "Measuring Competence? Exploring Firm Effects in
Pharmaceutical Research," Strategic Management Journal, 1994, 15, pp. 63-84. Ichniowski, Casey, Prennushi, Giovanna, and Shaw, Kathryn. "The Effects of Human Resource
Management Practices on Productivity." American Economic Review, May 1997, 87(3) pp. 291-313. Landers, Renee, James Rebitzer and Lowell Taylor. "Rat Race Redux: Adverse Selection in the
Determination of Work Hours in Law Firms." American Economic Review, June 1996, 86 (2),
pp. 329-348. Lazear, Edward, "Performance Pay and Productivity," NBER Working Paper 5672, 1996. Levine, David I. "Fairness, Markets, and Ability to Pay: Evidence from Compensation
Executives." American Economic Review, December 1993, 93(4),pp. 1241-1259. Lodge, David Nice Work. New York: Penguin, 1990. Piore, Michael. "Qualitative Research Techniques in Economics." Administrative Science
Quarterly, December 1979, 24, pp. 560-569. Smith, Adam. An Inquiry into the Nature and Causes of the Wealth of Nations. New York: Modern
Library, 1985. Thomas, Robert. What Machines Can't Do. Berkeley: University of California Press, 1994. Yin, Robert. Case Study Research. Beverly Hills: Sage Publications, 1984. Zucker, Lynne, Darby, Michael, Brewer, M. "Intellectual human capital and the birth of US
biotechnology enterprises." American Economic Review March 1998, 88(1) pp.290-306. 1.
Economics Department, Case Western Reserve University, Cleveland OH, 44106. I thank Martin Feldstein and
Adam Jaffe for involving me in the NBER/Sloan program, and David Levine for extensive discussions. I also
received much useful advice from those cited herein, Zvi Griliches, Carol Heim, and Don Siegel, The quote is from
Yogi Berra (nd).