### Table 7: Robustness results.

2004

"... In PAGE 30: ...We present another set of robustness results in Table7 . In the first column, we include the number of firms on the other standard as a third endogenous variable.... In PAGE 30: ... We re-estimate the model in column 2 of Table 5 but construct the likelihood value from only the 1909 observations that did not adopt at all in July.27 The results appear in column 2 of Table7 . The results are very similar to those estimated using all of the observations.... ..."

### Table 2: CTS robustness results.

### Table 2: Robustness results by problem class

2004

"... In PAGE 16: ...Table 2: Robustness results by problem class In Table2 we summarize the number (# Opt) and percentage (% Opt) of problems for which each solver reported nding the optimal solution, discriminated by problem charac- teristics. On 7 problems Snopt terminates with the message \optimal, but the requested accuracy could not be achieved quot; which implies that Snopt was within a factor of 10 2 of satisfying the convergence conditions.... ..."

Cited by 9

### Table 2: Robustness results by problem class

2003

"... In PAGE 16: ...Table 2: Robustness results by problem class In Table2 we summarize the number (# Opt) and percentage (% Opt) of problems for which each solver reported nding the optimal solution, discriminated by problem charac- teristics. On 7 problems Snopt terminates with the message \optimal, but the requested accuracy could not be achieved quot; which implies that Snopt was within a factor of 10?2 of satisfying the convergence conditions.... ..."

### Table 3: Robustness results by problem size

2004

"... In PAGE 17: ... We nd this encouraging since many features of our software implementation can be improved, as discussed in the nal section of this paper. Next we compare in Table3 the robustness of the solvers based on problem size. Note the decrease in reliability of Slique as the problem size varies from medium (M) to large (L).... ..."

Cited by 9

### Table 3: Robustness results by problem size

2003

"... In PAGE 17: ... We nd this encouraging since many features of our software implementation can be improved, as discussed in the nal section of this paper. Next we compare in Table3 the robustness of the solvers based on problem size. Note the decrease in reliability of Slique as the problem size varies from medium (M) to large (L).... ..."

### Table 3: EPFL robustness results 1 .

### Table 5: UCL robustness results 2 .

### Table 2: Robustness results by problem class

2003

"... In PAGE 17: ... We use the value n + m to characterize a problem apos;s size where n is the number of variables and m is the number of general constraints (not including bounds on the variables). In Table2 we summarize the number (# Opt) and percentage (% Opt) of problems for which each solver reported nding the optimal solution, discriminated by problem charac- teristics. On 7 problems Snopt terminates with the message \optimal, but the requested accuracy could not be achieved quot; which implies that Snopt was within a factor of 10?2 of satisfying the convergence conditions.... ..."