Malmendier & Tate CEO Overconfidence

Variables necessary to replicate Malmendier & Tate Over confidence measures:

Ideal Variables:

• Value of Options
• Option grant date
• Year in which option was exercised
• Mean stock price in year of exercise
• Median stock price in year of exercise
• High stock price in year of exercise

• Malmendier & Tate use the 5th year after the option grant date (maybe we use 5th year after person becomes CEO because we haven't found the grant date variable) because across various packages it is the first year in which at least part of the option packages in the sample are exercisable.

• 67% In the money appears to mean that the market price has appreciated 67% from the strike price as opposed to 67% of a CEO's options being in the money at a given time.

We construct Measure 1 (for both 67% and 100% in the money during the fifth year) as

follows. We consider the subsample of CEOs who at least twice during the sample period

had options that were valued above the threshold during the fifth year (and therefore after

the vesting period). We then identify the first instance at which the CEO failed to exercise

such an option during or before the fifth year. From this point in time onward, we classify

the CEO as overconfident as long as he subsequently exhibited the behavior at least one more

time during his tenure as CEO. This last requirement lessens the probability that someone will

be misclassified as overconfident when in fact they had an instance of inside information, they

were pressured to hold a particular option by the board, they faced a particularly high tax

burden from exercise, or they simply made a mistake. Indeed, as we are mainly interested in

the “permanent” rather than “transitory” overconfidence level of a CEO, our measure targets

CEOs who “habitually” exercise options late.

Our Malmendier type measure

We will come up with a measure to rank all the CEOs based on their option exercise behavior. This measure should first replicate some of the features of the overconfidence measure suggested by Malmendier and Tate. It should also exhibit sufficient variation in our sample of CEOs. After we have successfully ranked CEOs, we simply identify the top x percentile as overconfident.

The following annually recorded variable will be part of our measure:

Estimated value of in-the-money unexercised exercisable options($) VUE Estimated value of in-the-money unexercised unexercisable options($) VUU Value realized on option exercise($) VE

For each period [math] t [/math], let [math] OC_t^i = \frac{VE}{VE + VUE}[/math] to be the preliminary measure of overconfidence of CEO [math] i [/math] in period [math]t [/math]. I was also wondering if there would be any justification for including the variable VUU in the denominator.

Note that for each CEO we'd have multiple observations for different time periods. Thus, let the [math]OC^i [/math] to be the mean of all the observations for CEO [math] i [/math]