Supporting Information file.

Supporting Information A: Material for replicating the tasks. Supporting Information B: Loyalty Measure. Fig. A: Multi-dimensional plot resulting from the exploration of linguistic loyalty using the questionnaire from (29). Each of the statements below (English translation) was rated on a seven step scale between the poles “completely agree” and “strongly disagree”. Black circles (cluster 1) = statements on dialects with negative connotations (e.g., “Dialect is vulgar.”), black triangles (cluster 4) = statements on dialects with positive connotations (e.g., “Dialect conveys a feeling of security.”), white triangles (cluster 2) = statements on standard German with positive connotations (e.g., “Standard German sounds elegant.”), white circles (cluster 3) = statements on standard German with negative connotations (e.g., “Standard German sounds stiff.”). Kruskal’s test = .081. Fig. B: Dialect intensity across the speech samples. Fig. A shows the results of a formal linguistic test comparing the dialect intensity (i.e. regional accent in the given case) of our speech samples relative to standard language. The figure shows a strong and comparable deviation of both regional speech samples from standard language; the standard German samples are comparable with almost no indication of dialect features. The latter nicely illustrates the dual competence of our language informants. At the same time, we measure a small and insignificant difference in the dialect intensity between the two regional accent samples and between the two standard language samples suggesting that the difference does not influence language perception or affect the semantics of the text (32, 33). We calculate distance as number of micro-phonetic features like voicing, manner, or location of articulation that deviate from standard language divided by the overall number of words in the text (22). A value of d = 0 would suggest perfect compliance with standard language; a value of d = 1 means that, on average, one phonetically feature per word differs from standard language; very pronounced local dialects may have a score of d > 2 or even d > 3 (22, 32, 33). The difference between the regional accent samples is significant at p <. 001, the difference between the samples of the spoken standard language is not significant (cf. main text). Table A: The table shows panel regressions of tournament take-up where the outcome categories of revenue sharing and piece rate are pooled. Column 1 presents the results from a random effects panel regression on the choice of tournament and column 2–4 present mixed models. In column 3 we additionally control for the guessed rank of the EPs. Finally, in column 4 we add a full set of controls. Standard errors in parentheses; *p <. 10, **p <. 05, ***p <. 001. Table B: Linguistic loyalty and payment regime choice: Splitting the sample by loyalty measures, we observe that EPs with a high dialect loyalty (Panel A) choose tournament significantly more often when perceiving the distant Bavarian accent and significantly less often when perceiving the Thuringian accent. The same holds for EPs with standard German loyalty (Panel B), though the effect is less pronounced than it is in the case of dialect loyalty. For those EPs who have a low loyalty for dialects or standard German we do not find any effects. We report relative risk ratios; t statistics in parentheses. *p <. 05, **p <. 01 ***p <. 001
(DOCX)