Fig. 7: The impact of lag∆COVID-19 cases on ∆Fraud cases for high vs. low-fraud awareness subsamples.

This figure corresponds to Table 4. The x-axis is Lag ∆ Covid-19 cases, and the y-axis is ∆ Fraud cases. The line of Low-Fraud Awareness is drawn as follows. First, the product of the minimum of Lag ∆ Covid-19 cases for the subsample of low-fraud awareness times the coefficient on Lag ∆ Covid-19 cases for the same subsample is obtained. For the control variables, we obtain the products of the means of the variables for the subsample times their corresponding coefficient. The summation of all the products, which is the value of predicted ∆ Fraud cases based on the minimum of Lag ∆ Covid-19 cases for the subsample of low-fraud awareness, is then obtained. Second, the value of predicted ∆ Fraud cases based on the maximum of Lag ∆ Covid-19 cases for the subsample of low-fraud awareness is obtained in the same way as the minimum above, except that we use the product of the maximum of Lag ∆ Covid-19 cases times the coefficient on Lag ∆ Covid-19 cases. Third, we connect the two points between the values at minimum (min) and maximum (max) to form the line. The line of High-Fraud Awareness is formed the same way as the line of Low-Fraud Awareness, except that the coefficients and the values of the variables for the subsample of high-fraud awareness (instead of low-fraud awareness) are used (the coefficient on Lag ∆ Covid-19 cases is set to zero for the high-fraud awareness subsample because it is statistically insignificantly different from zero).