Authors:
John-Paul Ferguson – Stanford Graduate School of Business
Sharique Hasan – Stanford Graduate School of Business

Interviewers:
Yoonjin Choi – Columbia Business School
Mitali Banerjee – Columbia Business School

Article link: http://asq.sagepub.com/content/58/2/233

Question 1. How did you come to work on this project? Could you tell us more about finding the data? And how you arrived at the questions?

The back story with the data is straightforward. Sharique went to graduate school at CMU with an IAS officer, from whom he learned both some details of the service and about the public records of IAS-officer careers. The back story with the theory is a little more convoluted. John-Paul went to graduate school at MIT and then came to Stanford, probably the two places where the most debate about the benefits and penalties of category-spanning, particularly along the cognitive dimension, has been waged. We first started talking almost immediately after Sharique joined the faculty. At this time (circa 2010), a few people in the field were starting to present research projects that were kind of “horse races” between the supposed penalties of category-spanning and the supposed benefits of brokerage. We thought that the best way to reconcile these two ideas—a way that felt consistent with the types of processes that the authors of the theories described—was that yes, brokerage should give advantages, but that legitimacy was a logical and temporal precondition for brokerage to be exercised.

In a conversation about this, John-Paul described wanting a setting where actors were legitimated and where there was then a subsequent performance hurdle that had to be cleared—and, to make things more difficult still, there had to be no survivor bias or self selection! Without missing a beat, Sharique said, “I think I might know of some data where that happens.” And we were off and running.

All that sounds well and good, but notice that our initial predictions were different. Consistent with research on category-spanning, we thought that specialization would be a benefit at the early stage. Consistent with research on brokerage, we thought that diversification would be a benefit at the later stage. Imagine our chagrin when we found that specialization was rewarded at both stages!

Some beers followed. During that discussion, we had an idea: these theories effectively treat candidates as static, in the sense that their specialization is constant whenever they’re measured. Yet people aren’t static in their careers: their level of specialization is constantly changing. Because specialists are rewarded early on, they are given more opportunities to advance their career, and chances to deepen their skills. Thus they might eventually have the chance to branch out, but by that point the investment in a specific set of skills might be overwhelming, and they would prefer to continue doing one thing well. This would mean that, in the later stage, you would still see rewards for specializing—but only if people kept doing what they had been doing before the second promotion.

That was the key insight—because it suggested a new test. If our reasoning were correct, then we would find matches between what people had done before and after promotion later in their careers, but not earlier on. The ultimate form of the paper flowed from there.

Question 2. You do not lay out your hypothesis in the conventional way. It makes the reading more organic. Did you debate this decision? Why did you decide in its favor?

In all honesty, we did not have a large number of hypotheses, and it seemed overblown to structure the theory section so as to introduce just one or two. More importantly, we think that the paper’s contribution comes less from a startling or counter-intuitive set of hypotheses and more from finding better data and constructing better tests for an existing set of hypotheses that have proven very hard to study. So, we decided to spend more space on the setting and the tests than on the hypotheses as such.

Question 3. Table 5 (page 15) in the article suggests that specialization does not lead to a better match for early career officers except in “Finance”. Could this suggest something about the skills or relationships needed in finance? Also, the results in this table suggests that early career experience in the “Commerce” field decreases the likelihood of match in all but the “Personnel” field. This is interesting given the analysis in the appendix which suggests that “Commerce” might be a high-status field. Could a possible explanation have to do with the nature of skills/relationship gained or not gained by early career officers?

We do not think that there is something unique or special about finance here. It is the only significant coefficient on the diagonal, but we only showed the five most common specializations. There are a few other significant coefficients if you consider the full list of specializations, but we saw no pattern in what was significant and what was not, or indeed in what was positive and what was negative. This was in stark contrast to the same type of models examined in the later stages (summarized in table 7), where virtually everything on the diagonal was positive and significant.

As for the big, negative coefficient on Commerce that you point out, we think that this just reflects that specializing in Commerce is more rare. It is the rarest of the five we show in the table, for example. A handful of postings elsewhere after promotion for Commerce people would have a disproportionately large effect.

Question 4. We were intrigued by the results in table 3 (page 11), where early career female IAS officers are more likely to be posted to the “Center”. For female officers later in their career, the effect is positive but not significant. Could you tell us more about these effects? Is it specific to the IAS context which guarantees lifetime employment? Or perhaps it is suggestive of cross-cultural or institutional differences in outcomes for men and women?

We didn’t read too much into this. There are fewer women in the IAS, and their under-representation gets worse the farther back in time you go. We think the smaller number of women in the risk set in the later stages makes it harder to get precise estimates of the coefficient.

If we were to speculate about a real difference, we think one explanation might be that, for whatever reason, it has been harder for women to enter the IAS than it has been for men. This would imply that the underlying quality of a new female IAS officer might be higher than a male in her cohort. But once we pass through the Centre-posting stage, we have filtered out the relatively lower-quality male officers, and thus the quality of the men and women at risk of empanelment could well be comparable.

Question 5. You mention “guarantee of lifetime employment” as an idiosyncratic feature of your context. Do you see this as a scope condition for your findings? In general, what are the key scope conditions to bear in mind as we try to extend these results to other organizational contexts?

We would probably turn the point around, and say that doing research on career specialization is usually extremely hard, because most types of career data—surveys, administrative records from a single organization, and even career histories of more public figures—suffer from pretty massive survivor bias. After someone is fired, for example, they disappear from the records, and you usually cannot find out where they went afterward. Even if you know where they went, that organization might not give you comparable data. The result is that the people you observe tend to be more successful than average, and thus unrepresentative.

So we agree that lifetime employment is an idiosyncratic feature in our case, but we also think that it is a fundamentally important feature, insofar as it eliminates the problems of survivor bias. We think that that trade-off is worth making. Notice that, in our research setting, we find that specialization has career benefits even in the long term. This is not what prior studies have found. For example, the Zuckerman et al. typecasting piece said that the benefits of specialization fade away, but because unsuccessful actors exit from their population, we can’t know whether that fading away is real or driven by survivor bias. In other words, we do not think that lifetime employment is a theoretical scope condition on our study. Instead, it is an important empirical condition for finding unbiased results.

In terms of general scope conditions, an important one here and in other settings is that the hiring manager doesn’t know the prospective employee very well. This matters a lot because, following on the typecasting argument, we presume that the hiring manager focuses on the candidate’s observable and career history and not much more. If social networks play a large role in hiring, or selection happens based on a formal exam or the like, we would expect that signals from the career history would matter less. For example, actuaries and CPAs get promoted based on passing standardized tests at certain stages of their careers; and Bechky’s work for example documents how important social connections are to moving up the social hierarchy among film crews. These would not be ideal settings to test the theory.