Every academic has an opinion about what makes a good department. Surprisingly, there are few hard studies quantifying this precisely, although possible implications for an optimal design of education and research policies are numerous. Aghion et al. (2010) is an example of the general recent concern about the optimal design and governance of universities.
In our recent work, we try to start filling this gap and study the effect on individual publication records of a large set of department characteristics (Bosquet and Combes 2013).
- We use an exhaustive panel of academics working in French economics departments between 1990 and 2008 and their matched quality-adjusted publications in EconLit – the largest publication data set in economics.
- We control for many individual time-varying characteristics, individual fixed effects and reverse causality.
Features of French data
If using French data can be questionable for generalisation purposes, it bears some precious advantages when it comes to addressing issues surrounding endogenous location choices.
The problem is that individual publication records may drive department choice, so the correlation need not imply a causal effect. In France initial affiliation certainly relates to individual publications (as in most countries). This is captured by the individual fixed effect in our specification. However subsequent moves in France are not driven by individual performance. They are driven by either friendship connections or personal/family reasons.
This idiosyncratic feature of French academics is due to the fact that the civil-servant salary academics get is almost flat across universities. The most frequent way of becoming a full professor is via a national contest that allocates winners to departments in a largely random way. Combining these factors with an instrumentation strategy allows us to avoid using a natural experiment such as the dismissal of scientists in Nazi Germany used by Waldinger (2012).
Three components of individual productivity
We decompose individual productivity into three components:
- The probability to publish.
- The number of publications.
- The average quality of these publications.
We show that some variables have different effects from one productivity dimension to another, which means that the optimal strategy for an individual or a department depends on which dimension is being targeted.
For instance, some departments may be constrained in terms of the characteristics they can directly influence, and so may prefer to increase the proportion of publishing authors in the department. Others may consider specialising, i.e., allocating some people fully to teaching and administration to improve the quantity and quality of papers published by the others. Also, given their characteristics in terms of size and specialisation, some departments can find it easier to target the quantity of publications rather than the quality, or vice versa. These possibilities underline the importance of separately characterising the local design that would maximise each dimension of departments’ research productivity.
Large localised externalities in academic research
We first analyse the question about what makes an individual productive, i.e. is it her own abilities or the ‘firm’ in which she works? Applied to science, this translates into whether academics publish more because they have better abilities (gender, age, type of position, or any other individual characteristic possibly unobserved) and a more rewarding publication strategy (research field, number and location of co-authors) or because they are located in departments that provide a better local environment with stronger externalities?
Contrary to some recent results (Kim et al. 2009, Waldinger 2012), we find that location is an important determinant of the individual quantity and quality of publications. It represents at least half of the explanatory power of individual characteristics, which implies that local externalities in publication do matter.
We then move to the dual question of the extent to which more productive ‘firms’ simply attract more productive employees or generate more productive environments. Translated to academia the issue is whether good departments are either those where highly productive academics locate or those that generate more externalities.
- Even if we exhibit the presence of some spatial sorting of academics, the most productive academics being located in the best departments, we find that local externalities and the composition of departments equally matter when explaining the ranking of departments.
Departments’ characteristics that make researchers more productive
Finally, and most importantly for the optimal design of firms or institutions, what are the channels of local externalities?
- Size increases publications but not by much.
The impact of the size of the local economy on local productivity is one of the most studied questions in economic geography. We consider here the size of the department defined as its number of academics, which is at least partly in the hands of the department head, the university or the central government (in many European countries for instance).
We find that department size has a positive and significant impact on publication quantity (per academic), but has no effect on the other dimensions of research productivity (probability to publish and average quality). On top of that, it has a rather weak explanatory power – see Figure 1 for the relationship between department size and publication quantity (net of individual effects).
- Proximity to other economic departments does not matter either.
This contrasts with economic geography, where city size and market access are by far the largest local determinants of productivity.
Figure 1. Department size and publication quantity
- Publication performance heterogeneity is bad for publications.
Heterogeneity among researchers in terms of publication performance has a large, negative explanatory power. Figure 2 plots the strong negative relationship between department heterogeneity and the average publication quality (net of individual effects). The impact is similar for the probability to publish or the number of papers published. Increasing heterogeneity by 25% induces a decrease of 28% in the average publication quality, of 19% in the number of publications and of 16% in the probability to publish, everything else being equal.
Figure 2. Department heterogeneity and publication quality
- Field diversity is good for publications
The second department characteristic that has the highest explanatory power of individual publication performance is the diversity of the department in terms of research fields (within economics). Figure 3 emphasises that diversity is highly positively correlated with publication average quality (net of individual effects). Increasing diversity by 50% increases the average publication quality by 36.3%, quantity by 24.0% and the probability to publish by 17.3%.
Figure 3. Field diversity and publication quality
Finally, other department characteristics have interesting properties.
- Contrary to common intuition, more students per academic do not reduce publication performance.
- Women, older academics, stars in the department and co-authors in foreign institutions all have a positive externality impact on each academic's individual outcome.
All these results and others are detailed in in our paper should shed some first light on the optimal design of higher education and research policies.
Aghion, Philippe, Mathias Dewatripont, Caroline M. Hoxby, Andreu Mas-Colell, and André Sapir (2010), “The governance and performance of research universities: Evidence from Europe and the US”, Economic Policy 25(61):7–59.
Bosquet, Clément and Pierre-Philippe Combes (2013), “Do Large Departments make Academics more Productive? Agglomeration and Peer Effects in Research”, CEPR Discussion Paper 9401.
Kim, Han E., Adair Morse, and Luigi Zingales (2009). Are elite universities losing their competitive edge?, Journal of Financial Economics 93:353–381.
Waldinger, Fabian (2012), “Peer effects in science – Evidence from the dismissal of scientists in Nazi Germany”, Review of Economic Studies 79(2):838–861.