It is well established that insider trading can forecast returns. Firms whose shares have been intensively sold (bought) by insiders tend to underperform (overperform) benchmarks in subsequent periods.22 If some 1980s insiders forecasted poor performance for their firms, we might expect them to have looked for ways to keep the shareholders from firing them, either through voting or takeovers. In this case, weak shareholder rights would be a symptom of insiders' superior information, but would not necessarily be the cause of the poor performance in the subsequent decade. To study this possibility, we use data collected by Thomson Financial from the required SEC insider-trading filings. For each firm in our sample, we sum all (split-adjusted) open-market
transactions for all insiders in each year, with purchases entering positively and sales entering negatively. We then normalize this sum by shares outstanding at the beginning of the year to arrive at a "Net Purchases" measure for each firm in each year. If insiders put new provisions in place when they forecast poor performance, then we would expect Net Purchases to be negatively correlated with G.We employ two regression specifications. First, we estimate OLS regressions of Net Purchases on G (or a Democracy dummy),BM, and log of size. For some firm-years, the Net Purchase measure is dominated by one large transaction. While large transactions might have information content, they might also reflect liquidity or rebalancing needs. In an OLS regression, firms with large outliers will dominate. Thus, we also estimate ordered logit gressions on the same OLS regressors, in which the de pendent variable is equal to one if Net Purchases is positive, zero