Parametric and Nonparametric Market Power Tests:
An Empirical Investigation
Corinna M. Noelke*
Department of Resource Economics
217 Draper Hall
University of Massachusetts
Amherst, MA 01003
(413) 586-3497
(413) 586-3497 FAX
noelke@ipo.umass.edu
Kellie Curry Raper*
Department of Resource Economics
304 Draper Hall
University of Massachusetts
Amherst, MA 01003
(413) 545-5713
(413) 545-5853 FAX
raper@resecon.umass.edu
Selected Paper presented at the AAEA Annual Meetings
Salt Lake City, Utah August 4, 1998
Comments are welcome.
Abstract:
Parametric and nonparametric market power tests most commonly used to assess imperfectly competitive
behavior are identified. Monte Carlo experiments are used to evaluate the accuracy of eight
nonparametric market power tests. The results are compared to Raper, Love, and Shumway's
-
-
'
-
-
'
'
-
'
'
-
'
-
'
'
'
' '
'
' -
-
'
' '
'
'
'
'
' '
'
'
-
' '
''
''
-
(1)
Removing these outliers does not change the mean of the market power values significantly and does not
change the 'switched' values of the monopoly and monopsony market structures. Hence, Raper, Love,
and Shumway's monopoly market power test works to a certain extent; but again, the market structure
specification of the model is very important as the test detects some market power when market power
is instead being exerted from the opposite market.
Another modification of Love and Shumway's test is developed in this paper and restricts technical
change to be nonregressive. As no technical change parameters are estimated, a problem similar to
Ashenfelter and Sullivan's test is encountered since any measurement error or similar biases can only be
detected by the market power parameter. The results for the nonregressive monopsony market power test
point to such a problem, as all results are declared infeasible by the linear programming solver. Results
for the nonregressive monopoly market power test were not obtained because the test created problems
with the solver that could not be alleviated. The use of nonregressive technical change is reasonable;
however, it does not work in a linear programming formulation.
Parametric Market Power Tests
Raper, Love, and Shumway (1997) report mean values and standard deviations of market power
parameters over 1000 simulations for a Bresnahan-Lau type monopoly market power test, a monopsony
market power test, and FlexPower, using the same data set as this study. The 'true' values for 
for
Bresnahan-Lau type parametric market power tests are 1.0 for monopoly, 0.5 for Cournot duopoly, and
0.4046 for Stackelberg duopoly. 
should be equal to zero for all other market structure data. The
'true' values for
are 1.0 for monopsony, 0.5 for Cournot duopsony, and 0.3956 for Stackelberg
duopsony. For all other market structures, 
should be equal to zero.
Both the monopoly market power test and the monopsony market power test using the
Bresnahan-Lau approach perform remarkably well. FlexPower combines the two uni-lateral market
power tests into one test that does not assume a priori one side of the market to be perfectly competitive,
but allows for either or both sides of the market to have some degree of market power. FlexPower gives
results similar to the monopoly and monopsony tests when considering the significance of market power
estimates. Additionally, FlexPower is able to distinguish between perfect competitive and bilateral
monopoly data.
Conclusions
Knowledge of the degree of market power exertion is important in guiding antitrust and merger
policies. With the help of Raper, Love, and Shumway's -
'
'
'
'
'
'
'
''
'
'
'
-
" " - "
" - " " - "
"
- " "
- " "
- " - "
- " "
- " "
- "
"
"
"
"
"
" "
-
' '
' '
1. Data available upon request.
