Can More Money Reduce the Rate of Associate Attrition?

A recently published story in The American Lawyer asks the salient question, "Is Raising Salaries the Best Way to Retain Associates?"   It is always great to get an unsolicited $20,000 raise, but what if your real gripe is that you work too many hours and your physical health and family life are falling apart?  Or that there is no communication on partnership chances, making sticking around too risky?  It is possible that managing partners and executive committees do not know the actual price points--monetary and non-monetary--of their associates.

Some of the findings presented in my essay with David Zaring, "Young Associates in Trouble," suggest that these law firms may, in fact, be throwing good money after bad.  In the table below, which is based on data from the Am Law Midlevel Associate Survey (3rd, 4th, and 5th year lawyers), the dependent variable is likelihood on a 1-5 scale that the associate will remain with the firm during the next two years. 

The regression results tell a very straightforward and intuitive story:

• Carrots.  Midlevels are more likely to stick around if (a) they do interesting work or work on quality projects (which I suspect means career enhancing), (b) are given cues on partnership prospects, or (c) the firm has a large nonequity tier (this study explains why this is attractive to non-rainmaking lawyers).
• Sticks.  Midlevels are less likely to stick around when the hours are long and the environment is less family friendly.
• Irrelevant.  Controlling for all of these facts, higher salaries and bonuses appear to have no statistically relationship to Midlevel's self-reported likelihood of staying with the firm.

Comments

[William Henderson]

John,

You raise a good question. In fact, there is not a very strong correlation between compensation and interesting work or quality work. Multicollinearity is not typically a problem unless the correlations are > .7.

I also check the variance inflation factor (VIF) diagnostics. The usual rule of thumb is that VIF > 5 are problematic and may require re-specification of the model. The VIF were all lower than that.

[John Coates]

Interesting, and within some range, plausible. But I don't think your regression gets me there: what's the correlation between compensation and interestingness/quality of work? If it's high, as I expect, then multicollinearity makes inference about both difficult. As a thought experiment: suppose a firm scoring high on interestingness/quality of work cut salaries by 50% for incoming associates. Do we really think the associates would be as good, or that, conditional on taking the job in the first place, stick around as long?