Does Leadership Matter?

In our last post, we introduced Nelson Mandela’s challenge to political economy: incorporate the role of leadership and ideas into our theoretical and empirical investigations.

This is a bit like world peace. Pretty much everybody is in favor of it — or at least says they are in favor of it. But few know how to achieve it.

Of course it’s easy to criticize existing approaches because they don’t have leadership and a role for ideas independent of interest. But that’s also a little cheap.

Though some would disagree, we believe firmly that economics — and more generally social science — makes progress by being much more specific about how certain social phenomena can be modeled formally; by deriving new insights and perspectives that would not have been obvious without doing this modeling work; by devising new ways of measurement corresponding to these new concepts; and by testing hypotheses and new ideas systematically using state-of-the-art econometric and statistical methods.

So the devil is in the details.

Be that as it may, there is actually work in political economy and economics in general on these topics.

First, two related strands of work show that leaders do indeed matter.

The first is exemplified by a creative paper by Benjamin Jones and Benjamin Olken, “Do Leaders Matter?”. Focusing on changes in national leadership resulting from random leader death (where leader here means the person with executive power which could be president, prime minister or dictator), Jones and Olken find evidence that leaders do matter.

In particular, they test for the hypothesis that the growth rate of GDP is the same before and after the random leader death, and comfortably reject this hypothesis, suggesting that there is a change in growth as leaders die and are replaced by new ones.

Two things are particularly interesting about this paper.

First, they are testing whether a change in leadership is associated with any change in growth — a hypothesis much weaker than a change in leadership from somebody with certain characteristics (say low education) to other characteristics (say high education) increasing economic growth. This is for good reason. There doesn’t seem to be as strong patterns when it comes to leader characteristics (but see this paper and this paper).

Second and more interestingly, Jones and Olken show that this result is driven by changes in leadership in countries with weak and non-democratic political institutions. This says that the change from George W. Bush to Barack Obama, important though it was for many things, should not have changed the US growth rate, which is plausible.

This makes a lot of sense at some level. Good — what we would call inclusive — political institutions will allow less discretion for politicians to pursue policies that would be disastrous, and hence gyrations of economic growth should also be more limited.

It also implies that if you believe — as we firmly do of course — that institutions are central for understanding economic performance, there is no conflict with this view and one that shows that leaders do matter: leaders function in a framework created by institutions and their impact very much depends on and is modulated by these institutions (though of course they may in turn also shape the evolution of institutions as we will discuss later).

The second is a related line of work investigating the impact of Chief Executive Officers (CEOs) on company performance. An important paper that is the precursor to Jones and Olken’s work is Marianne Bertrand and Antoinette Schoar’s paper on “Managing with Style” which uses a related but in some ways richer empirical design exploiting the fact that, differently from national leaders, CEOs manage several major companies during their careers (or work as other top managers in such companies, which Bertrand and Schoar also incorporate into their sample). Thus one can estimate an econometric specification in which company performance can be decomposed, among other things, into “CEO effects” and “firm effects” (formally, this means the estimation of a panel data regression in which there are both CEO and firm fixed effects).

Bertrand and Schoar also find that these CEO effects are statistically important — which is the equivalent of Jones and Olken’s finding that the hypothesis that growth before and after random leader death being equal can be rejected.

In addition, using the data available for practices under the reign of different CEOs, they can have a first attempt at investigating why different CEOs matter. In particular, they show that dividend policy, investment policy, cost-cutting strategies, financial policies and merger and acquisition decisions are particularly sensitive to who holds the reins of the company.

This conclusion is corroborated by more recent work by Nick Bloom and John Van Reenen which shows huge differences in management practices across and within countries, and provides evidence suggesting that these are likely to be quite important for firm performance.

These studies make considerable progress in showing that leadership at the company or the country level matters, and also putting some new ideas on the table related to in what ways leadership may matter.

Of course, there are many open questions. For example, at the national level, the death of a leader may matter not because of his character or vision, but because there are different “rent seeking coalitions” associated with different leaders (think of the rent seeking coalition in Egypt before and after Mubarak).

When one leader dies, this leads to the dissolution of one rent seeking coalition and the formation of another. This will naturally have an impact on economic performance even if the two rent seeking coalitions have exactly the same preferences over economic growth and investment in different sectors of the economy. Add to this the possibility that different rent seeking coalitions are supported by and will favor different businesses in sectors of the economy, it becomes quite possible that leadership transitions can be associated with very significant changes in economic growth without any of the notions about “leadership” we normally think about — particularly when being inspired by Nelson Mandela — being important.

At the company level, the same problem can occur, though this is easy to put to rest with the evidence that many important company practices do change significantly with leadership changes.

More important at the company level is the fact that leadership changes may be endogenous, partly occurring when the company already wants to change course in terms of some of these policies (this wasn’t as much of an issue at the country level because of Jones and Olken’s focus on random leader deaths).

Though there is probably some of this going on, it seems unlikely that the presence of this sort of endogenous CEO change by itself can explain all of Bertrand and Schoar’s results. Also, even if it could bias their point estimates, the qualitative results shouldn’t be affected, since the only reason why there would be a systematic change from one type of CEO to another when the company wants to change course is that some CEOs are better with or more likely to adopt some types of practices or will implement them more effectively, thus providing another type of support to importance of leadership.

This is of course just scratching the surface. Much more is needed to advance our understanding of how leadership matters at the national or the company level. We also need to study when there are leader transitions and also perhaps whether, when and how existing leaders change their strategies and approaches because of changing circumstances.

So there is much to be done here.

The same is true when it comes to the modeling of leadership, which we will discuss in our next post.

PrintView Printer Friendly Version

EmailEmail Article to Friend

« Modeling Leadership | Main | Nelson Mandela, Leadership and Ideas »