Appendix to “ The Role of Context and Team Play in Cross - Game Learning ”

Context * Crossover Cycle 1 -.564 (.242) -.564 (.275) -.938 (.451) -1.007 (.647) Abstract Context * Crossover Cycle 2 -.385 (.232) -.385 (.395) -.620 (.355) -.687 (.588)Context * Crossover Cycle 2 -.385 (.232) -.385 (.395) -.620 (.355) -.687 (.588) Abstract Context * Crossover Cycle 3 .152 (.308) .152 (.367) .207 (.463) .068 (.639)Context * Crossover Cycle 3 .152 (.308) .152 (.367) .207 (.463) .068 (.639) Meaningful Context * Crossover Cycle 1 .727 (.244) .727 (.194) 1.127 (.394) 1.338 (.659) Meaningful Context * Crossover Cycle 2 1.214 (.253) 1.214 (.412) 2.294 (.406) 2.454 (.658) Meaningful Context * Crossover Cycle 3 .467 (.318) .467 (.388) .845 (.466) .845 (.641) 2x2 * Crossover Cycle 1 .698 (.232) .698 (.155) 1.482 (.407) 1.425 (.629) 2x2 * Crossover Cycle 2 .444 (.287) .444 (.287) 1.133 (.478) 1.031 (.648) 2x2 * Crossover Cycle 3 1.322 (.336) 1.322 (.279) 2.116 (.556) 1.925 (.671) Log Likelihood -1414.62 -1414.62 -1053.68 -1046.41 * statistically significant at the 10% level ** statistically significant at the 5% level *** statistically significant at the 1% level Note: Parameter estimates are identical in Models 1 and 2, as clustering only affects the size of the standard errors. Table A.4 Experiment 1: Comparing Outcomes Against Within Treatment Controls (With Entry Rate Controls and Alternative Controls for Individual and Session Effects) Variable Model 1 Model 2 Model 3 Model 4 Individual and Session Effects Clustering Player Level Clustering Session Level Random Effects Player Level Nested Random Effects (Player/Session) # Clusters 419 32 419 419/32 Abstract Context * Crossover Cycle 1 -700 (.277) -700 (.288) -1.078 (.499) -1.030 (.552)Context * Crossover Cycle 1 -700 (.277) -700 (.288) -1.078 (.499) -1.030 (.552) Abstract Context * Crossover Cycle 2 -.613 (.226) -.613 (.310) -.911 (.405) -.891 (.471)Context * Crossover Cycle 2 -.613 (.226) -.613 (.310) -.911 (.405) -.891 (.471) Abstract Context * Crossover Cycle 3 .296 (.307) .296 (.326) .162 (.431) .067 (.490)Context * Crossover Cycle 3 .296 (.307) .296 (.326) .162 (.431) .067 (.490) Meaningful Context * Crossover Cycle 1 .333 (.258) .333 (.311) .547 (.386) .864 (.583) Meaningful Context * Crossover Cycle 2 .800 (.260) .800 (.319) 1.690 (.388) 1.982 (.587) Meaningful Context * Crossover Cycle 3 .691 (.312) .691 (.301) .979 (.418) .934 (.548) 2x2 * Crossover Cycle 1 .385 (.242) .385 (.192) 1.041 (.528) 1.000 (.550) 2x2 * Crossover Cycle 2 .114 (.293) .114 (.289) .630 (.563) .555 (.576) 2x2 * Crossover Cycle 3 1.016 (.341) 1.016 (.304) 1.531 (.530) 1.406 (.600) Entry Rate Differential 1.139 *** (.204) 1.139 (.243) 1.444 (.274) 1.302 (.298) Log Likelihood -1364.20 -1364.20 -1039.52 -1037.18 * statistically significant at the 10% level ** statistically significant at the 5% level *** statistically significant at the 1% level Note: Parameter estimates are identical in Models 1 and 2, as clustering only affects the size of the standard errors. Table A.5 Experiment 2: Crossover from Quantity to Price Game (Alternative Controls for Individual and Session Effects) Variable Model 1 Model 2 Model 3 Model 4 Individual and Session Effects Clustering Player Level Clustering Session Level Random Effects Player Level Nested Random Effects (Player/Session) # Clusters 275 21 275 275/21 Constant -.015 (.193) -.015 (.142) -.552 (.283) -.166 (.301) Cycles 2 – 4 .361 ** (.163) .361 (.296) .781 (.325) .841 (.359) Cycles 3 – 4 .053 (.168) .053 (.186) .118 (.332) .114 (.347) Cycle 4 .162 (.158) .162 (.310) .669 (.347) .718 (.331) 1 x 1 * Meaningful Context * Crossover Cycles 1 – 4 -.277 (.265) -.277 (.215) -.829 (.401) -.999 (.438) 1 x 1 * Meaningful Context * Crossover Cycles 2 – 4 -.346 (.221) -.346 (.303) -.547 (.429) -.611 (.453) 1 x 1 * Meaningful Context * Crossover Cycles 3 – 4 .472 (.247) .472 (.256) 1.056 (.449) 1.017 (.469) 1 x 1 * Meaningful Context * Crossover Cycle 4 -.079 (.200) -.079 (.332) -.208 (.470) -.221 (.451) 1 x 1 * Abstract Context * Control Cycles 1 – 4 -.208 (.274) -.208 (.420) -.276 (.490) -.307 (.595 1 x 1 * Abstract Context * Control Cycles 2 – 4 -.254 (.238) -.254 (.326) -.056 (.458) -.143 (.540) 1 x 1 * Abstract Context * Control Cycles 3 – 4 .261 (.231) .261 (.275) .775 (.504) .875 (.523) 1 x 1 * Abstract Context * Control Cycle 4 .071 (.233) .071 (.372) -.200 (.505) -.291 (.489) Log Likelihood -1188.89 -1188.89 -701.62 -696.56 * statistically significant at the 10% level ** statistically significant at the 5% level *** statistically significant at the 1% level Notes: Parameter estimates are identical in Models 1 & 2 – clustering only affects the size of the standard errors. Table A.6 Experiment 2: Crossover from Quantity to Price Game (With Entry Rate Controls and Alternative Controls for Individual and Session Effects) Variable Model 1 Model 2 Model 3 Model 4 Individual and Session Effects Clustering Player Level Clustering Session Level Random Effects Player Level Nested Random Effects (Player/Session) # Clusters 275 21 275 275/21 Constant -1.468 *** (.357) -1.468 (.338) -1.714 (.507) -.944 (.546) Cycles 2 – 4 .825 *** (.182) .825 (.277) 1.126 (.418) 1.091 (.376) Cycles 3 – 4 -.093 (.185) -.093 (.286) -.072 (.463) -.021 (.278) Cycle 4 -.230 (.214) -.230 (.478) .464 (.363) .491 (.315) 1 x 1 * Meaningful Context * Crossover Cycles 1 – 4 -.187 (.279) -.187 (.300) -.546 (.425) -.815 (.474) 1 x 1 * Meaningful Context * Crossover Cycles 2 – 4 -.492 (.223) -.492 (.315) -.642 (.481) -.642 (.453) 1 x 1 * Meaningful Context * Crossover Cycles 3 – 4 .275 (.268) .275 (.314) .846 (.516) .874 (.403) 1 x 1 * Meaningful Context * Crossover Cycle 4 .109 (.230) .109 (.492) -.089 (.436) -.115 (.426) 1 x 1 * Abstract Context * Control Cycles 1 – 4 .035 (.283) .035 (.325) .156 (.507) -.356 (.680) 1 x 1 * Abstract Context * Control Cycles 2 – 4 -.331 (.251) -.331 (.353) -.196 (.586) -.230 (.529) 1 x 1 * Abstract Context * Control Cycles 3 – 4 .062 (.263) .062 (.406) .748 (.626) .780 (.467) 1 x 1 * Abstract Context * Control Cycle 4 .158 (.262) .158 (.484) -.226 (.512) -.283 (.463) Entry Rate Differential 2.233 *** (.438) 2.233 (.489) 1.606 (.516) .146 (.064) Log Likelihood -1119.11 -1119.11 -696.65 -694.26 * statistically significant at the 10% level ** statistically significant at the 5% level *** statistically significant at the 1% level Notes: Parameter estimates are identical in Models 1 & 2 – clustering only affects the size of the standard errors. The data set includes a large number of observations from experienced subject sessions with a crossover between the high and low cost entrant games. Only observations where the low cost entrant game was played are included. These provide a source of additional independent clusters based on experienced play of the low cost entrant game. This data is dummied out of the regressions – in other words the estimates reported above do not reflect this data. None of these parameter estimates are reported above as these are not of direct interest. The base for the regressions are quantity-price crossover sessions with abstract context. Table A.7 Probit Regressions, Crossover from Quantity to Price Game Standard Errors Corrected for Clustering at the Individual Level Dependent Variable: Strategic Choice by MLs Location Pittsburgh Ohio State Variable Model 1 Model 2 Model 1 Model 2 Constant -.210 (.160) -.194 (.158) -.015 (.194) -.136 (.209)