The issues: Methodology and substance
• Studies in all three areas face similar methodological issues
• Number of observations
• Statistical reliability
• Possible selection/bias/distortion
• Risk of undue focus on what is measurable
• Further substantive issues
• Nature of harm (price vs nonprice)
• Short vs longer term effects
• Non-economic effects
The case of mergers
• Merger studies confront all these issues
• Ideal body of evidence would be studies of a large random samples of “policy-relevant mergers” that measure several important outcomes
• In reality, largest body of evidence comes from merger retrospectives
• These are before-and-after studies of effects: Difference-in-differences
• Isolate merger effects by controlling for other factors possibly causing observed effects
Merger retrospectives and meta-analyses
• Merger retrospectives are now common but they have limitations
• Not necessarily random
• Tend to focus on one or two measurable effects, especially price
• Most examine one merger at a time
• To capture the collective implication of this body of evidence, best technique is a meta-analysis of retrospectives
• Meta-analysis compiles and synthesizes results of multiple studies
• Uses common format, permitting summary statistics across all studies
• My meta-analysis from 2015 appears still to be as large and robust as any
• That study compiled all published U.S. merger retrospectives
• 60 published studies covered 42 distinct mergers in 20 industries
Findings of studies
• Key analysis reported that prices rose in more than 80% of studied mergers
• Overall average price rise was 7.2 %
• Among those 80 %, average price rise was 9.5%, ranging up to 29.4%
• Confirmed by my second meta-analysis of studies of multiple mergers
• Collectively covered thousands of mergers and reported similar effects
• Results further confirmed by later studies cited in Background Note:
• Olsen et al report effects on large number of mergers (not all at enforcement margin) between 5% and 7%
• Ashenfelter-Weinberg study of 5 mergers finds 3-7% price rise
• Two studies of consumer/package goods show smaller effects, but these are all mergers, not just policy-relevant ones
Closer look at evidence: First result
• Overall, best evidence is that price effects of “policy-relevant mergers” average between 3% and 7%
• This comes from hundreds of studies and thousands of mergers, conducted by numerous researchers
• These are after controlling for other possible causes
• This is first important result
Closer look at evidence: First result
• Overall, best evidence is that price effects of “policy-relevant mergers” average between 3% and 7%
• This comes from hundreds of studies and thousands of mergers, conducted by numerous researchers
• These are after controlling for other possible causes
• This is first important result
• But the same average can arise from different patterns of actual data
Further look at evidence: Second result
• Second important result concerns this possible variation in outcomes
• If all measured effects are near average, that single number is sufficient
• But if data indicate a wider range of effects, the use of simple average for all cases will make mistakes that can be costly
• My study found substantial variation
• Estimates of price increases range between 0.7% to 29.4%
• Outcomes include some price decreases, as large as 4.9%
• Similarly wide range of effects are reported in other studies
• Possible policy measures to address this variation
• Median of data may be better than mean since it captures “typical” case
• Could use lower quartile to ensure “conservative” estimate, although in this context “conservative” means explicit underestimating
• Or could use different metrics depending on industry, characteristics
• Mean vs median:
• My study found average price increase of 9.5% but median was 5.9%
• Lowest quartile
• In my study, this was 3.5%, below both mean and median
• Different measures according to characteristics known to affect magnitude
• By industry:
• Olsen finds largest price distortion in healthcare, airlines
• My study found largest distortion in hospitals, publishing, airlines
• Thus, for such mergers, policy might consider larger percent
• By characteristics:
• Largest price effects arise in the most concentrated industries
• Effects most pronounced where entry barriers are high/moderate
• Thus, for mergers in high concentration industries, one could use larger percent, and even greater where barriers are high/moderate
Third result and conclusions
• Third result is the important reminder that price distortions are only one aspect of competitive effects of mergers
• Adverse effects on quality
• Innovation effects, especially over time
• Many fewer studies of these other effects
• They are more difficult to measure, or subject to significant lags in their outcomes
• Recommendations
• Conduct more merger retrospectives
• Greater focus on nonprice outcomes