Difference between revisions of "Shue (2011)"

From edegan.com
Jump to navigation Jump to search
Line 2: Line 2:
 
|Has page=Shue (2011)
 
|Has page=Shue (2011)
 
|Has title=
 
|Has title=
|Has author=
+
|Has author=Shue
 
|Has year=2011
 
|Has year=2011
 
|In journal=
 
|In journal=

Revision as of 13:23, 29 September 2020

Article
Has bibtex key
Has article title
Has author Shue
Has year 2011
In journal
In volume
In number
Has pages
Has publisher
© edegan.com, 2016

Return to BPP Field Exam Papers 2012

Author's Hypothesis

Executive social interactions are important determinants of managerial decision making and firm policies. The author tests to see if executive and firm outcomes are more similar among section peers than among class peers using data from HBS MBA students.

How the Author Tested Hypothesis

The author adds additional structure of a linear in means model. The model is specified as:

[math]Y_{isc}=\theta\bar Y_{sc} + \phi\bar v_{sc} + \alpha_{sc} + \rho_{isc}[/math]

The individual outcomes [math]Y_{isc}[/math] either compensation or acquisitions are effected by both mean group outcomes [math]\bar Y_{sc}[/math] and mean group fundamentals [math]\bar v_{sc}[/math]

In the baseline model, he does not differentiate between the two types of peer effects in calculating the peer elasticity.

1.)The author creates a pairs distance metric. Hypothesis is that the mean absolute distance in outcomes between section peers should be less than the distance in outcomes between class peers. This is estimated using a 2 stage procedure as described on page 17. Notice that SE need to be adjusted as individuals appear multiple times in construction the pairs.

2.)The author also creates an excess variance metric. This tests the hypothesis that peer influences should also reduce variance within peer groups relative to across groups.

Trade offs between methods (1) and (2) Excess variance relies on familiar ANOVA model and results are easily comparable to previous work on peer effects. Also, estimation of elasticity does not require assumption of normality of individual fundamentals as in pairs distance measure. However, the pairs distance measure is more robust to outliers. because it relies on distance not distance squared as variance would.

3.) Finally, the author uses an "exogenous" shock or alumni reunions to test if peer effects in compensation and acquisitions are driven by contemporaneous interactions or past interactions.

4.) He also adds a robustness check in Pay for Friends Luck

What Tests Achieved

He finds strong evidence of peer effects in executive compensation and acquisitions. Under the linear in means model he estimates a lower bound of elasticity of individual outcomes of 10%-20%.

In the year following a reunion, section peers are around 15% more similar than class peers and the implied elasticity is 30%. Equality in the distance ratio in the year after reunions compared to other years can be rejected at the ten percent level for all outcomes except total compensation. Altogether, tests using alumni relations seem to show that peer effects in compensation and acquisition are driven by contemporaneous interactions. Pay for friends luck tests show that section peers are 6-10% more similar than class peers even when the peers compensation is due to lucky shocks in his industry.

How might the tests be improved upon

I am slightly worried about the exogenous nature of reunions. Reunions occur every 5 years, and while its likely that peers do not keep in touch outside of the reunion, I am worried about some other sort of convergence effect. Perhaps, executives keep fairly close tabs on their close friends who are from their section group, and if they have less acquisitions or something work to make this up in the lead up to a reunion, so its not so much the reunion as the mechanism, as a keeping up with the Johnson's effect.

What is an alternate empirical strategy