The role of relative versus absolute time differences between interviews Notes: The figure shows estimates of the autocorrelation based on equation (2). The black (gray) dots show the autocorrelation between the vote given to the candidate interviewed in t and the candidate interviewed in $t−1(t−2)$⁠, depending on the time between the end of the interview in $t−1(t−2)$ and the start of the interview in t. Note that the autocorrelation with $t−2$ and $t−1$ are estimated on two different subsets of interviews. The cut-off at 45 minutes is chosen as the minimum time distance between $t−2$ and t. Dashed lines show 95% confidence intervals. Standard errors are clustered at the workshop/candidate pool level (N = 312/N = 63)