Empirical research in an increasingly concentrated industrial environment: discussion.

These three papers argue persuasively that to understand the welfare impacts of changes in agricultural markets we need more and better data. In this brief comment, I summarize the papers individually, synthesize the papers' suggestions for improved data collection, and provide an example of data collection in a concentrated industry.

Summary of Papers

All three of the papers argue that increased concentration gives individual firms more latitude to shape their environment. Since the firms are no longer price-takers who are tiny relative to their markets, their strategic choices have a large impact on social welfare. However, academics and policy-makers lack data to understand these developments. The papers each give a cogent example of these problems.

Fernandez-Cornejo and Just argue that changes in agricultural input markets (due to increases in both intellectual property protection and the rise of biotechnology) have led to increased productivity (due to increased R&D), but also to increased concentration (due to increased patenting). The authors present several examples of how arcane details of regulations have important anti-competitive effects. However, data limitations prevent economists from evaluating the net welfare effects of these developments.

Perloff and Denbaly have a similar message for retail markets: questions about the efficacy of increased grocery store concentration cannot be answered with existing data. They also point out that existing data do not allow serious study of important food safety and health questions. Their comments about the difficulty of tracking the food supply are particularly timely in light of recent revelations about the prevalence of unsafe Chinese seafood and chemicals in the U.S. food supply.

Huth, Ligon, and Dimitri argue that lack of data prevents studying another important change in agricultural markets: the rise of contracts between farmers and intermediaries. They also show the value of field research in designing effective data collection efforts. For example, an older distinction between production contracts and marketing contracts is not useful, because contracts currently often combine elements of both types. Thus, it is difficult for respondents to answer questions that ask which type of contract they have, and for researchers to interpret the results.

These papers make valuable contributions in two ways. First, they summarize important changes in agricultural markets, especially increased concentration. Second, they describe in detail important policy questions raised by these changes, and the data that would be necessary to begin to answer them. As Perloff and Denbaly note, data is nonrival in its use, and hence will be underprovided by markets. Thus there is an important role for nonmarket actors such as government and academia.

Implications for Data Collection

What form should this data collection take? The papers suggest four principles:

1. Those who benefit from government programs have an obligation to provide data to help evaluate the programs, and for other social purposes (for a compendium of good practices in data collection and program evaluation, see Levine 2006). Fernandez-Cornejo and Just, and Huth, Ligon, and Dimitri both suggest more detailed surveys of farmers to get at (a) patterns of seed and pesticide use and (b) the nature of contracting. Perloff and Denbaly suggest surveying food-stamp recipients to understand linkages between food consumption and health. Data-for-assistance is an important principle, but it should be applied to powerful aid recipients (such as giant companies that receive patents) as well as to farmers and food-stamp recipients.

2. An understanding of current practice is necessary to collect meaningful data.

3. Data collection should meet rigorous standards: samples should be representative, and response rates meaningful. Statistics on these metrics should be disclosed.

4. Government and academic researchers should partner with each other to collect new data, and to negotiate lower prices for access to commercial data sources. Investments in linking existing data sources (such as linking data on food consumption with data on health) can be particularly productive.

I propose three additional points to think about:

1. Experiments with random assignment to treatment and control groups can be a powerful way of measuring program effectiveness where there are large numbers of potential participants.

2. The proliferation of expensive private data sources potentially creates a vicious circle for data collection--and not just in agriculture. As free government data becomes less relevant to firms' needs, they will be more willing to pay for private data. Once they have paid for private data they may become more resistant to participating in government data collection and to paying taxes to fund it. This resistance will make public data seem even less relevant, even though (as the papers point out) private data is gathered with much less rigor and described with much less care than is public data.

3. Partnerships among universities can remedy gaps in data collection in concentrated industries. For example, the International Motor Vehicle Program, one of twenty-six "industry centers" in a network organized by the Sloan Foundation, carries out multiple studies of the auto industry. In doing this work, we found that interviews are a key tool, because they help us to (a) learn what questions to ask in surveys, (b) understand parties' objectives and constraints, (c) gather vivid images, and (d) interpret what we find in surveys (Helper 2000). We also found that firms will participate in well-designed surveys that address issues of interest to them. However, while university-based surveys can be useful pilots, they don't substitute for government-collected data, because academics cannot compel response, and do not have the resources to generate continuous data series.

As a nation we are not spending nearly enough on data to answer the important questions raised by these papers. To take just one example, suppose sub-optimal policies lead to a 1% higher price of corn (due either to monopoly pricing or lack of innovation due to excess competition). This price increase would cost $263 million per year--20% more than the entire combined annual budgets of the USDA's National Agricultural Statistics Service and Economic Research Service.

References

Helper, S. 2000. "Economists and Field Research: 'You Can Observe a Lot Just by Watching'." The American Economic Review 90:228-32.

Levine, D. 2006. "Building Learning into the Global Aid Industry." http://faculty.haas.berkeley.edu/ levine/#_Development_Economics, accessed August 6, 2007.

Sloan Foundation. 2007. "Programs: Standard of Living and Economic Performance." http:// sloan.org/programs/IndustryStudies.shtml, accessed August 6, 2007.

Susan Helper is Professor, Department of Economics, Case Western Reserve University.

This article was presented in a principal paper session at the AAEA annual meeting (Portland, OR, July 2007). The articles in these sessions are not subjected to the journal's standard refereeing process.