Financial attention and the disposition effect, controlling for investor sophistication 

Contexts in source publication

Context 1

... estimated coefficients and standard errors from this extended specification are pro- vided in Table 4, and a visualization of the model's implications for the disposition effect are shown in Figure 2. The magnitude of the estimated probabilities, and their difference appear largely unaffected by the inclusion of these sophistication controls. A reason for this outcome is that financial attention may capture a component of investor skill and sophisti- cation unaccounted for by such other proxies. However, we also find that only the average wealth invested at the broker has a significantly negative coefficient on its interaction with gain ijt . The coefficient on the interaction effect between portfolio diversification and gain ijt is even significantly positive. Nevertheless, the latter result may also reflect the ongoing discussion on whether underdiversification results from an informational advantage (see for instance Ivkovi´c Ivkovi´c et al. (2008), Van Nieuwerburgh and Veldkamp (2010) and Korniotis and Kumar ...

Context 2

... * indicate significance at the 1, 5, and 10 percent level, respectively. Standard errors are clustered at the investor and stock level and are reported in parenthesis. Table 6 reports coefficients and standard errors from the logistic regression model described in Section 4, where the dependent variable takes the value of one if an investor decides to sell a stock position on a day when he decides to sell at least some stock, and 0 if otherwise. Observations are obtained for a subsample, which excludes observations that are possibly due to the rebalancing of the stock portfolio. Coefficients marked with ***, **, and * indicate significance at the 1, 5, and 10 percent level, respectively. Standard errors are clustered at the investor and stock level and are reported in parenthesis. Table 7 reports coefficients and standard errors from the logistic regression model described in Section 4, where the dependent variable takes the value of one if an investor decides to sell a stock position on a day when he decides to sell at least some stock, and 0 if otherwise. Observations are obtained for a subsample, which only includes observations that are the result of a market order sale. Coefficients marked with ***, **, and * indicate significance at the 1, 5, and 10 percent level, respectively. Standard errors are clustered at the investor and stock level and are reported in parenthesis. Table 8 reports coefficients and standard errors from the logistic regression model described in Section 4, where the dependent variable takes the value of one if an investor decides to sell a stock position on a day when he decides to sell at least some stock, and 0 if otherwise. Observations are obtained for the 2014 subsample, where each attention decile is constructed based upon investors' portfolio monitoring behavior in 2014. Coefficients marked with ***, **, and * indicate significance at the 1, 5, and 10 percent level, respectively. Standard errors are clustered at the investor and stock level and are reported in parenthesis. Table 9 reports coefficients and standard errors from the logistic regression model described in Section 4, where the dependent variable takes the value of one if an investor decides to sell a stock position on a day when he decides to sell at least some stock, and 0 if otherwise. Observations are obtained for the 2015 subsample, where each attention decile is constructed based upon investors' portfolio monitoring behavior in 2015. Coefficients marked with ***, **, and * indicate significance at the 1, 5, and 10 percent level, respectively. Standard errors are clustered at the investor and stock level and are reported in parenthesis. Table 10 reports coefficients and standard errors from the logistic regression model described in Section 4, where the dependent variable takes the value of one if an investor decides to sell a stock position on a day when he decides to sell at least some stock, and 0 if otherwise. Observations are obtained for the 2016 subsample, where each attention decile is constructed based upon investors' portfolio monitoring behavior in 2016. Coefficients marked with ***, **, and * indicate significance at the 1, 5, and 10 percent level, respectively. Standard errors are clustered at the investor and stock level and are reported in parenthesis. Table 11 reports coefficients and standard errors from an extended logistic regression model, containing additional interaction effects with a dummy variable, Extreme ijt , which is equal to 1 if the cumulative return since purchase exceeds 30% in absolute value. The dependent variable takes the value of one if an investor sells a stock position on a day when he sells at least some stock, and 0 if otherwise. Observations are obtained for 20,709 unique clients, who traded in common equity at the Belgian discount broker between January 2014 and December 2016. Coefficients marked with ***, **, and * indicate significance at the 1, 5, and 10 percent level, respectively. Standard errors are clustered at the investor and stock level and are reported in parenthesis. The top panel of Figure 1 reports the average predicted probabilities of selling a winning or a losing stock position, as a function of investors' financial attention, expressed in deciles. The probabilities are implied from the logistic regression described in Section 4, where the dependent variable takes the value of 1 if an investor sells a stock position on a day when he sells at least some stock, and 0 if otherwise. The bottom panel reports the change in the disposition effect, as a function of financial attention. The disposition effect at each decile of attention is calculated as the average partial effect of a dummy variable, taking the value of 1 if the stock position was trading at a gain in the investor's portfolio, and 0 if otherwise. Dotted lines represent 95% confidence intervals. Standard errors are clustered at the investor and stock level. The top panel of Figure 2 reports the average predicted probabilities of selling a winning or a losing stock position, as a function of investors' financial attention, expressed in deciles. The probabilities are implied from an extended logistic regression model with additional controls for investor sophistication and heterogeneity. The dependent variable takes the value of one if an investor sells a stock position on a day when he sells at least some stock, and 0 if otherwise. The bottom panel reports the change in the disposition effect, as a function of financial attention. The disposition effect at each decile of attention is calculated as the average partial effect of a dummy variable, taking the value of 1, if the stock position was trading at a gain in the investors portfolio, and 0 if otherwise. Dotted lines represent 95% confidence intervals. Standard errors are clustered at the investor and stock level. The top panel of Figure 3 reports the average predicted probabilities of selling a winning or a losing stock position, as a function of an investor's financial attention, expressed in deciles. The probabilities are implied from the logistic regression described in Section 4, where the dependent variable takes the value of 1 if an investor sells a stock position on a day when he sells at least some stock, and 0 if otherwise. The bottom panel reports the change in the disposition effect, as a function of financial attention. The disposition effect at each decile of attention is calculated as the average partial effect of a dummy variable, taking the value of 1 if the stock position was trading at a gain in the investors portfolio, and 0 if otherwise. Results are calculated for a subsample of clients, who only traded in equities throughout the sample period. Dotted lines represent 95% confidence intervals. Standard errors are clustered at the investor and stock level. The top panel of Figure 4 reports the average predicted probabilities of selling a winning or a losing stock position, as a function of an investor's financial attention, expressed in deciles. The probabilities are implied from the logistic regression described in Section 4, where the dependent variable takes the value of 1 if an investor sells a stock position on a day when he sells at least some stock, and 0 if otherwise. The bottom panel reports the change in the disposition effect, as a function of financial attention. The disposition effect at each decile of attention is calculated as the average partial effect of a dummy variable, taking the value of 1 if the stock position was trading at a gain in the investors portfolio, and 0 if otherwise. Results are calculated for a subsample, excluding observations that are possibly due to rebalancing efforts. Dotted lines represent 95% confidence intervals. Standard errors are clustered at the investor and stock level. The top panel of Figure 5 reports the average predicted probabilities of selling a winning or a losing stock position, as a function of an investor's financial attention, expressed in deciles. The probabilities are implied from the logistic regression described in Section 4, where the dependent variable takes the value of 1 if an investor sells a stock position on a day when he sells at least some stock, and 0 if otherwise. The bottom panel reports the change in the disposition effect, as a function of financial attention. The disposition effect at each decile of attention is calculated as the average partial effect of a dummy variable, taking the value of 1 if the stock position was trading at a gain in the investors portfolio, and 0 if otherwise. Results are calculated for a subsample of observations, which are the result of a market sale order. Dotted lines represent 95% confidence intervals. Standard errors are clustered at the investor and stock level. No. Days Logged-in (Deciles) Figure 6 reports the disposition effect at each decile of attention, calculated as the average partial effect of a dummy variable, taking the value of 1 if the stock position was trading at a gain in the investors portfolio, and 0 if otherwise. The results are implied from the logistic regression described in Section 4, where the dependent variable takes the value of 1 if an investor sells a stock position on a day when he sells at least some stock, and 0 if otherwise. Results are calculated for subsamples of observations spanning each year, and where each attention decile is determined based upon investors' portfolio monitoring behavior in that year. Dotted lines represent 95% confidence intervals. Standard errors are clustered at the investor and stock level. No. Days Logged-in (Deciles) Figure 7 reports the average predicted probabilities of selling a stock position that is trading at a moderate or an extreme return in the investor's portfolio, as a function of an investor's financial attention, expressed in deciles. The top panel reports the probabilities for a stock position that is trading at a profit, ...