Table A.2: Descriptive statistics of inputs and outputs (in logs)
|
Variable |
Mean |
St. Dev |
CV |
Min |
Q1 |
Median |
Q3 |
Max |
|
Output |
16.89 |
149 |
8.85 |
12.36 |
15.78 |
1671 |
17.85 |
25.29 |
|
Material inputs |
15.77 |
1.84 |
11.64 |
0 |
14.58 |
15.71 |
16.96 |
24.91 |
|
Labor compensation |
15.76 |
1.36 |
8.64 |
10.29 |
14.73 |
15.55 |
16.58 |
23.97 |
|
Energy consumption |
12.65 |
1.72 |
13.61 |
0.61 |
11.39 |
12.47 |
13.75 |
22.25 |
|
Capital |
14.07 |
1.51 |
10.70 |
8.94 |
13.02 |
13.94 |
15.00 |
22.50 |
|
External services |
13.38 |
2.01 |
15.00 |
2.87 |
11.99 |
13.36 |
14.73 |
22.34 |
|
Other inputs |
14.40 |
1.76 |
12.22 |
7.93 |
13.14 |
14.28 |
15.54 |
23.38 |
Notes: CV is the coefficient of variation; Q1 and Q3 are lower and upper quantiles.
Such average values of yearly depreciation result in a considerably higher estimate of the output
elasticity of capital. We are aware that using a proxy variable instead of a direct measure of the
capital stock input could be of concern. However, even with such a crude proxy based on the
tax depreciations for the capital input, we obtain estimates of the elasticity of capital that appear
to be quite reasonable.
The sample contains a number of observations with extreme values (see maximum and
minimum columns in Table A.2) that proved to have a considerable impact on the estimated
parameters of the production function and led to implausible results. Therefore, we exclude
such "outliers" from the analysis when the cost for a certain input category in relation to gross
value added is less than the lowest 0.1 percent and the highest 99.9 percent. In total, these
excluded cases plus firms with zero values for at least one input category (the major part of
excluded cases) account for about 10 percent of all observations. We find that the exclusion of
these extreme cases leads to a considerable improvement of robustness and plausibility of the
estimation results for the production function.
29