First, the magnitudes of partial R2s for the effect of industry affiliation (Table 7) clearly
reinforce the results of the previous section and therefore support Hypothesis 1. Despite the
fact that in absolute terms, the partial R2s of industry affiliation for the best and worst perform-
ers are larger than for the medium performers, in relative terms, industry affiliation provides
approximately 40 percent (more than 50 percent for the 1999-2005 period) of the explanatory
power of the models. Thus, Hypothesis 1 holds true irrespective of the firms’ level of productive
efficiency.
Second, within the subgroup of medium performers, the size effects are similar to those
(Table8) observed for the entire sample (Table 5) for both periods, 1992-2005 and 1999-2005.
Moreover, in this subgroup, larger firms are, again, less efficient than their smaller counterparts.
For the worst performers, however, size has no explanatory power. In the group of best perform-
ers, the size effects have only 0.02 percent explanatory power and lead us to reject Hypothesis 3
for the worst and best performing firms.
Third, location effects are notably different across the three subgroups. Location effects
are not statistically significant for the group of best performers. However, they are pronounced
for the worst performers in period 1992-2005, but, oddly, not significant for the period 1999-
2005. The parameter estimates of the district dummies reflect the average efficiency of the
firms located in the respective district. Though in the beginning of the 1990s, firms in East
Germany have been rather inefficient as a result of the transition of the former socialist regime,
this clear East versus West separation in the efficiency of districts can not be found for the
later period of 1999-2005. Rather, there is a mixture of East and West German districts among
the least and most efficient locations, indicating that locational effects are not solely due to
East or West German regional differences but might be caused by other (nonobserved) reasons.
Thus, Hypothesi 2 is supported with regard to medium performers, but not for worst and best
performers.
Fourth, the results for the medium performer subgroup also confirm Hypotheses 4 and 5. A
heterogenous picture emerges for the best and worst performing firms (Table 8). For example,
the quota of material inputs has a positive impact for the worst and medium performers but is
not significant for the best performers. The quota of external services has a negative impact on
efficiency for worst performers but is not significant for the two other groups. However, exter-
nal contract work is conducive to efficiency for the best performers. Thus, in addition to the
relatively low explanatory power of outsourcing activities the evidence on the direction of ef-
fects for efficiency are ambiguous. Likewise, for the worst and best performers, R&D intensity
is statistically insignificant. Both partial R2s as well as the coefficient are statistically signifi-
cant only for medium performers. Thus, surprisingly, R&D does not explain any statistically
significant variation of productive efficiency at the two ends of the efficiency distribution.
Overall, three effects are responsible for most of the explanatory power: (i) industry, (ii)
size, and (iii) location. All other factors, both firm-specific and environmental, yield statistically
19