Conceptual Framework

We describe a simple model that formalizes our intuition on the impact of educated leaders and guides our empirical analysis. We highlight two important mechanisms through which an educated leader could have an impact. Both of the mechanisms we suggest are unobservable and untestable directly in our empirical framework. However, we flesh out the implications of the two mechanisms and test the implications. The model considers only the action of an elected leader and do not explicitly take into account the actions of voters or other political actors.

We assume that there is a limited amount of resource in terms of influence, effort, or time avail-able to each leader. A leader can use the resource to act uponndifferent areas such as public infrastructure, health, education, extracting revenue to increase one’s own wealth, raising funds for next campaign etc. The leader has to decide on how to distribute the total amount of resource on different activities with the goal of maximizing own utility. Leaders differ among each other on basis of the level of education (G) – which for instance can be whether college graduate or not. If the leader putse_ilevel of resource in theith area (wherei={1, ..., n}), then the resulting output is given by the following function:

137

Ai(e_i, G) = _i(G)ei

which can be viewed as an achievement of the leader in the context of this particular area. The choice of this specific functional form is to keep the illustration simple. _i(G)is a parameter that captures the competency of the leader in improving outcomei, and it depends on leader’s own education levelG. It is plausible to assume that ^@A_@eⁱ

i = i(G)> 0. If the outcome of interest is education, then the achievement might be increasing enrollment rate, learning levels of students or improving school quality. We observe the achievements but do not observe the actual resource allocated or effort exerted by the leader across different areas directly.

The leader derives utility from achievement on various items through several channels. Firstly, achievements reflect a leader’s performance which in turn affects the probability of winning the next election. Secondly, leaders also get direct utility from achievement on areas that create a lasting legacy or give them a sense of satisfaction and fulfillment. Therefore the leader’s utility is a function of his/her achievements on thendifferent items. For ease of exposition we assume a log-linear form of the utility function, and express the leader’s problem as follows:

maximize Pn

i=1⇥i(G)log[A_i(e_i, G)]

{e1, e2, ..., en}

subject to Pn

i=1e_i=1

The preference parameter⇥_i(G)reflects how much weightage is given to achievement in theith area by a leader withGlevel of education. We assume⇥i(G)>0andPn

i=1⇥i(G) =1. The constraint indicates that there is an upper bound to the total resource available and it has been normalized to 1.

Solving the first order conditions of this problem we find that the leader’s optimal allocation of resource towards theith item is given by:ei⇤=⇥i(G). Therefore the optimal level of achievement is expressed as:

A_i(e^⇤_i, G) =A^⇤_i(G) = _i(G)⇥_i(G)

So the optimal achievement onith item is directly proportional to competency and preference of the leader towards this particular item. Since both competency and preference parameters are non-negative, therefore achievement will increase withGif at least one of competency ( i(G)) and preference (⇥i(G)) increases withG. In other words, an educated leader yields better outcome in an area if he/she has higher competency or higher preference towards that area as compared to a

138

non-educated leader.

A competent leader or leader who understands a particular issue might be able to achieve better outcome in the area ithan a less competent or less knowledgeable leader for the same level of effort. Leaders might also differ in their achievements in sectoribecause they obtain more utility from achievements in the sector above the ones obtained by other leaders due to their preferences.

While we cannot empirically separate out preference from competence, we test the claim in our empirical model that achievements of educated leaders are higher than less-educated leaders.

Besley and Reynal-Querol (2011) argue that good leaders might differ based on competency and prefer broad-based policies and advocate public goods and infrastructure that generate wide ben-efits. They argue that educated leaders are better citizens, have higher talent and higher concern for social welfare. Elementary education is a basic public good whose widespread provision can lead to improvement in welfare of citizens. If the arguments made by Besley and Reynal-Querol (2011) or Atkinson et al. (2016) are valid then educated politicians would lead to better quality of elementary education as they are more competent and would prefer to invest their resources in education due to its impact on long term welfare of citizens. Even if educated leaders do not prefer to invest in education, their higher competency by itself should on average lead to better education outcomes.

While acknowledging these arguments Carnes and Lupu (2016) argue that the link between for-mal education, competence and leadership is not as straightforward. It is possible to gain human capital and skills required for being an effective leader without formal education. Formal educa-tion does not just reflect human capital but also the privileges of being able to obtain educaeduca-tion.

Societal restrictions and economic hardship might hinder a talented person from obtaining formal education while a less talented person might be able to obtain formal education. There still exists widespread discrimination, though declining, in obtaining even elementary education by women and individuals from lower castes in India.

Carnes and Lupu (2016) also argue that human capital obtained through formal education alone does not necessarily improve quality of leaders. Crucial factors like character, personality, ability to listen and understand people’s grievances and other leadership qualities are not the focus of formal education and could be obtained without it. Even basic reading and arithmetic skills could be obtained outside of formal schooling. In instances where specialized skills are required leaders could rely on qualified bureaucracy to help formulate solutions. Several studies have also shown that most qualified and those with highest grades are not necessarily most successful people in the

139

society (Gottesman and Morey, 2006; Clotfelter et al., 2007). Several studies in the 1960s and 70s and recent followups by Carnes (2012, 2013) which studied relationship between education level of leaders and their attitudes, choices and decisions when in office found no difference in behavior between more and less educated politicians (see Carnes and Lupu (2016) for details).

The empirical evidence on impact of education level of the leader on policies is thin and mixed.

Dreher et al. (2009) find that professional and highly educated leaders are more likely to imple-ment market-liberalizing reforms. On the other hand, Carnes and Lupu (2016) show that across contexts and wide range of outcome indicators politicians with a college degree perform the same or worse than non-college graduate politicians. Educated leaders at national level in a cross-country database, legislative leaders in the US and local municipal leaders in Brazil all perform no better than non-educated leaders. This holds across a range of outcomes including economic growth, inequality, social unrest, interstate conflict, unemployment, inflation, reelection, legislative pro-ductivity, and corruption.

Most of the studies which found a positive impact of education qualification of the leader analyze leaders at the national level but the impact of education of the leader might be different at lower administrative levels (except a few studies such as Martinez-Bravo (2014)). Leaders at district or constituency level work more as “fixers” or lobbyists for their constituents and have smaller role to play in broader policy making (Chopra, 1996). Analyzing the activities of elected state con-stituency representatives across Indian states, Jensenius and Suryanarayan (2015) find that state leaders spend most time in their constituencies rather than debating legislation in state assemblies.

The time spent by leaders debating legislation and making policies has also been declining over time – from 45 days per year in 1967 to about 34 days per year – and most legislation is passed without much debate. Politicians tend to spend most of their time in their constituencies addressing their constituents’ complaints, attending social functions, being part of local government bodies, helping individuals in accessing various government schemes, lobbying the district and state ad-ministration to implement their favored schemes and use their networks to attract investment to their constituency (Chopra, 1996; Jensenius, 2015). Formal education might be helpful but is not essential to gain the skills required for this set of tasks.

140