and maximizing Equation (1) using the parameters in the experiment, we obtain a PLAH of 0.38, which implies a Prob(Investigation) of 0.55.
In expectation we would observe 38 percent of the low assets disclosed as high in the I-cell and an average PLAH setting of 0.38 in the M-cell.
Ex ante, one can do no better than use a PLAH of 38 percent.
The more frequently the seller lies (e.g., the higher the PLAH is set), the lower is the buyer's expected value.
Regardless of whether the M-cell sellers choose to leave PLAH at a constant or shift it between 0 and 100, they can mimic the I-cell sellers directly or in fraud level.
Those sellers are predicted to act like their I-cell counterparts by switching between 0 and 100 for the PLAH. Alternatively, they could use the computerized mechanism to remove the responsibility for the disclosure fraud from themselves to the information system.
(9) For cell M, the average of each of the respective 20-round groups of PLAH values for each seller is computed.
The average PLAH in the I-cell was significantly different from the predicted 38 percent rate (p < 0.003).
The results show that the average of the PLAH for all sellers was more than 50 percent for only three (M-cell) or four (I-cell) subjects in the last 20 rounds.
The PLAH was averaged across all sellers (cell M) or sellers with a low-quality asset (cell I), giving 10 observations for each round subgroup.
A separate regression was calculated for each cell using the average per-round PLAH within each cell as the dependent variable and the round as the independent variable.