The Effectiveness of Competition Policy

1
The Effectiveness of Competition Policy

• Robert W. Crandall
• The Brookings Institution
• Lear Conference
• Rome, Italy
• 25-26 June, 2009

2
To Paraphrase Shakespeare (and John Kwoka)

• I come not to praise competition policy, nor to
  attack it but to analyze it
• In 2003, Cliff Winston and I offered the
  following thought
• we envision that economists may identify
  cases where antitrust policy has improved
  consumer welfare.
• However, to quote Fred Jenny (28/03/08, Ankara)
• The economic literature contains little
  systematic analysis of overall effects of
  competition law enforcement.

3
Buccirossi, et. al. ,Competition Policy and
Productivity Growth An Empirical Assessment
Addresses Jennys Lament

• Impressive piece of work that deserves careful
  reading
• Nevertheless, I have serious reservations about
  some aspects of the paper, including
• The choice of the dependent variable
• The specification of the basic regression
  equation estimated with pooled cross-section,
  annual time series data
• The assumption that EU countries competition
  policy have effects only in their own countries
• The measurement of RD
• The lack of a variable to capture cyclical
  influences
• The limited breadth of the industries in which
  competition policy has a significant effect
• The lack of any significant effects of
  competition policy in service industries

4
Empirical Method

• Authors use pooled cross-section, time-series
  regressions with fixed effects on annual
  industry-level data.
• TFP growth in year t is related to measures of
  Human Capital, RD Intensity, Import Penetration,
  the Technology Gap, Product-Market Regulation,
  and the Competition Policy Indexes (CPIs) in year
  t-1(with the exception of the Product
  Market Regulation variable).
• But year-to-year TFP growth is quite volatile and
  is heavily affected by the business cycle, yet
  there is no industry-specific cyclical variable
  in the equation(s).
• Because of this cyclical volatility, most
  analyses of TFP growth look at longer-term
  trends, not year-to-year variations.

5
The Choice of Dependent Variable Why TFP?

• The effect of competition policy should be
  registered in increases in output in
  non-competitive industries reductions elsewhere
• The authors settle on TFP (total factor
  productivity), suggesting that competition policy
  increases innovation and productive efficiency.
• However, competition policy could have major
  effects on allocative efficiency without having
  much effect on TFP.
• If competition policy has a beneficial effect in
  certain industries, that effect should be
  reflected in higher output growth. Why not see if
  the Competition Policy Indexes (CPIs) affect
  output growth in these industries?
• Alternatively, the authors could examine the
  effect of the CPIs on economic growth in the
  spirit of Dutz and Hayri (1999).

6
Cross-Country Evidence on Trends in TFP

• The current study focuses on 1995-2004 and uses
  data from the nine EU countries, the United
  States, Canada and Japan
• van Ark, et.al. (2008), using much of the same
  data as in the current study, have shown that
  between 1980-95 and 1995-2004
• TFP in the European Union declined dramatically
  from 0.9 per year to just 0.3 per year
• But TFP in the United States nearly trebled from
  0.5 per year to 1.4 per year.
• Labor productivity growth in market services rose
  from 1.5 to 3.2 per year in the U.S., but fell
  from 1.6 to 0.9 per year in the EU and 0.9 per
  year in the EU
• Market services were responsible for 60 of the
  growth in U.S. labor productivity over this
  period

7
The Competition Policy Variables

• Extremely detailed attempt to capture the
  differences in the policy environment across
  countries
• Too complicated to discuss in detail here, but I
  note that the Institutional and Enforcement
  Indexes reflect the underlying policy structure,
  not its execution
• For instance, the average attendee at this
  conference would be surprised to find that the
  Competition Policy Index (CPI) for the U.S. fell
  slightly during the second Clinton term, but rose
  somewhat during the first (George W.) Bush term!
• The CPIs tell us nothing about the intensity of
  antitrust enforcement nor whether it is
  effectively targeted on anti-competitive
  practices

8
Competition Policy Indexes North America and
Japan
9
Competition Policy Indexes -- Large EU Countries
10
Competition Policy Indexes -- Small EU Countries
11
Major Changes in CPIs Drive the Empirical Results

• I show the charts for the aggregate CPIs because
  it appears that changes in the CPIs in a few
  countries drive the empirical results, given the
  inclusion of country-industry dummy variables.
• With one exception the United Kingdom beginning
  in 1999 the largest changes occur in small EU
  countries.
• Note large increases in
• Czech Republic, beginning in 1995
• Hungary, beginning in 1999
• Netherlands, beginning in 2000
• More modest increases in Spain and France circa
  2000
• Note that increases in EU countries CPIs occurs
  at a time when EU is falling behind U.S. in TFP
  growth.

12
Further Problems with the Model

• Effects of CPIs are limited to industries in the
  Competition Authorities own countries despite
  the fact that 9 of the 12 countries are in the
  EU. The effects of competition policy in each EU
  country surely extend to the entire EU, or at
  least to a large subset of EU countries.
• Moreover, it is very difficult to believe that
  any change in the CPI would affect TFP with just
  a one-year lag
• RD variable is measured by annual RD spending
  divided by value added in the given industry
  rather than the stock of RD capital.
• Nevertheless, the RD variable is statistically
  significant when used, probably because it is
  picking up cyclical influences on TFP growth, but
  it is then dropped from all subsequent equations.
  Why?
• No overall goodness-of-fit statistics reported.
  What share of variance in TFP is explained by the
  model?

13
Significant Effects of CPIs Are Confined to a Few
Industries, Reflecting a Very Small Share of GDP

• Industries with significant coefficients (share
  of 2007 U.S. GDP)
• Agriculture, forestry, and fishing (1.2)
• Food processing -- (1.3)
  
• Paper, printing, and publishing -- (0.7)
• Petroleum and coal products -- (0.5)
• Rubber products -- (0.5)
• Non-metallic mineral products -- (0.4)
• Machinery --
  (0.9)
• Furniture --
  (0.3)
• Total with significant effects 5.8
• Industries with insignificant coefficients
• Professional and Business Services (12.2)
  Finance (7.9) Construction (4.4)
• Transport and Storage (2.9) Communications
  (3.2) Hotels and Restaurants (2.7)
• Electricity, Gas Water (2.0)
• Omitted

14
Any Corroborating Evidence in these Industries of
Increasing Competition?

• Food processing?
• Paper and printing?
• Petroleum refining?
• Nonmetallic minerals (cement, etc.)?
• Machinery?
• Rubber products (tires)?
• Furniture?
• If Competition Policy is effective in these
  industries, there should be some other evidence
  of entry, reductions in dominant positions,
  successful attacks on price-fixing conspiracies,
  etc.

15
In These Industries, U.S. TFP Growth Exhibits
Enormous Volatility, Even Over 5-Year
Periods(Average Annual Growth)
16
An Alternative Test The Effects of Competition
Policy on Overall Economic Growth

• As I mentioned earlier, the authors could test
  the effect of the CPIs on economic growth in the
  spirit of Dutz and Hayri (1999).
• GDP growth in the EU has consistently lagged
  behind the U.S. by about 25 percent since the mid
  1980s in large part because of less investment in
  ICT.
• Since 1999, the average CPI in the nine EU
  countries in the sample has risen relative to the
  CPI for the U.S.
• Although there is no evidence that EU growth has
  accelerated relative to U.S. growth since 1999,
  is it possible that after taking into account
  the other (exogenous) determinants of growth
  the increase in the EU CPIs has favorably
  affected EU growth?

17
Concluding Observations

• Buccirossi, et. al. , have given us a provocative
  paper that concludes that Competition Policy, as
  they measure it, has virtually immediate effects
  on productive efficiency.
• These effects apparently occur in a few EU
  countries and across only a few manufacturing
  industries, not in service industries.
• These results contrast with, but do not
  necessarily contradict, another analysis of the
  same data (van Ark, et. al. 2008) which finds
  that the EU lagged badly behind the U.S. in TFP
  growth in the same 1995-2004 period
• The results of this study might change
  substantially if the RD variable were properly
  constructed and used in the all of the relevant
  regression equations and if cyclical variables
  were included in the regressions
• Such changes might induce the authors to
  investigate the effects of more plausible lags in
  the effect of their CPIs on TFP, labor
  productivity, or economic growth