Difference-in-differences analysis in antitrust: A cautionary note

Antitrust Law Journal

Volumn 75, Issue 2, 2008, Pages 623-635

  Simpson, John ^a   Schmidt, David ^a  

^a FEDERAL TRADE COMMISSION   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 57049104620     PISSN: 00036056     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Review

References (23)

1. For a discussion of this methodology, see Jonathan B. Baker & Daniel L. Rubinfeld, Empirical Methods in Antitrust Litigation: Review and Critique, 1 AM. L. ECON. REV. 386 (1999);
2. Laurence Schumann, James D. Reitzes & Robert P. Rogers, In the Matter of Weyerhauser Company: The Use of a Hold-Separate Order in a Merger with Horizontal and Vertical Effects, 11 J. REG. ECON. 271 (1997). A variant on this methodology estimates the relationship between price and cost and demand factors and then forecasts price during the treatment period.
3. See Halbert White, Time-Series Estimation of the Effects of Natural Experiments, 135 J. ECONOMETRICS 527 (2005).
4. For examples of studies using this approach, see David M. Barton & Roger Sherman, The Price and Profit Effects of Horizontal Merger: A Case Study, 33 J. INDUS. ECON. 165 (1984);
5. Severin Borenstein, Airline Mergers, Airport Dominance, and Market Power, 80 AM. ECON. REV. 400 (1990);
6. Justine S. Hastings, Vertical Relationships and Competition in Retail Gasoline Markets: Empirical Evidence from Contract Changes in Southern California, 94 AM. ECON. REV. 317 (2004) [hereinafter Vertical Relationships and Competition];
7. Justine S. Hastings & Richard J. Gilbert, Market Power, Vertical Integration, and the Wholesale Price of Gasoline, 53 J. INDUS. ECON. 469 (2005);
8. E. Han Kim & Vijay Singal, Mergers and Market Power: Evidence from the Airline Industry, 83 AM. ECON. REV. 549 (1993);
9. Robin A. Prager & Timothy H. Hannan, Do Substantial Horizontal Mergers Generate Significant Price Effects? Evidence from the Banking Industry, 46 J. INDUS. ECON. 433 (1998);
10. John Simpson & Christopher T. Taylor, Do Gasoline Mergers Affect Consumer's Prices? The Marathon-Ashland and UDS Transaction, 51 J.L. ECON. (forthcoming February 2008);
11. Christopher T. Taylor & Daniel S. Hosken, The Economic Effects of the Marathon-Ashland Joint Venture: The Importance of Industry Supply Shocks and Vertical Market Structure, 55 J. INDUS. ECON. 419 (2007).
12. Making a similar point about DID estimators in a broader context, Abadie notes: "[I]t is well known that the DID estimator is based on strong identifying assumptions. In particular, the conventional DID estimator requires that, in the absence of the treatment, the average outcomes for the treated and control groups would have followed parallel paths over time." Alberto Abadie, Semiparametric Difference-in-Differences Estimators, 72 REV. ECON. STUD. 1, 1 (2005).
13. While a 20 percent cost shock may seem large, the control markets in at least one of the studies cited previously experienced price changes of this magnitude. See Hastings, Vertical Relationships and Competition, supra note 2, at 323 (examining price changes in gasoline markets in Southern California).
14. o) where T and C respresent the treatment and control markets respectively, and 1 and 0 represent the post-treatment and pre-treatment periods respectively.
15. This particular case is simplified to illustrate our main point.
16. While one might contend that the difference in the pre-treatment prices in the treatment and control markets might cause a judge or jury to question a comparison of price changes in the two types of markets, our main point would also hold in more complex examples where pre-treatment prices in the treatment and control markets were more similar. 0.5(0.75 - 0.75) + 0.5(0.8 - 0.75) = 0.025.
17. See Gregory J. Werden & Luke M. Froeb, The Effects of Mergers in Differentiated Products Industries: Logit Demand and Merger Policy, 10 J.L. ECON. & ORG. 407 (1994).
18. Throughout this example, the market price is measured as the market-share weighted average price across all firms using premerger shares.
19. The DID approach could also yield inaccurate results in this example if the treatment and control markets differed only in some other market characteristic. But the results would not be nearly as sensitive to differences in these other characteristics as they are to differences in the degree of substitutability.
20. The reason the merger effect is 9 percent, instead of the 7 percent reported by Werden and Froeb, is that in addition to the price change caused by the merger, this figure also accounts for the fact that the merged company in the treatment market passes through less of the cost decrease than its non-merged counterparts in the control market.
21. To put these differences in perspective, these parameters imply that the proportion of customers who would leave MCI due to a price increase that would switch to AT&T (the "diversion ratio") is 60 percent in the treatment market and 42 percent in the control market.
22. Another way to interpret this graph is that a merger simulation model calibrated to reflect the control market would incorrectly predict the effect of a merger in the treatment market by exactly the amount shown in the graph.
23. In other cases, the same uncertainty about market conditions that complicates the application of the DID approach may also complicate the application of other measurement techniques. In these cases, a researcher might prefer the DID approach because of its simplicity.