Should Politicians Control Civil Servants?

In his column for GURU, NES Professor Michele Valsecchi tells about his current research.

Broadly speaking, the theme of my current research is bureaucracy and its relations with politicians. The research question that I’m looking into is whether politicians should have control over bureaucrats. There are two different views or theories on this.

The first view is that modern bureaucracies should be insulated from politicians as much as possible. That’s because in the 19th century and before the civil service was dominated by patronage (or at least that’s the idea that many people have), which means hiring people just because they are loyal to you, for example, they are your friends or relatives. So, to avoid that, starting from the 19th century, bureaucracies in the US and elsewhere were being reformed on the broad principle that recruitment should be meritocratic and based on a competitive examination and that promotions should be based on rigid rules in order to avoid manipulation. Also, within this approach civil servants should not become political agents, i.e. they should refrain from being members of any political party, or from promoting any political side. In some countries, although not in all, it is regulated by law. This concept of insulating bureaucrats from politicians is based on the ideas of several sociologists, the most prominent of them is Max Weber, and this is the view I held as well.

But there is also another view as well. In democracies, politicians are supposed to be accountable to citizens. This happens primarily via voting. Now, people cannot observe exactly what politicians do, or how well they do it, but they can observe socio-economic outcomes (like GDP, quality of education or quality of health) and use those instead. The problem is that the latter are the result of many factors, including some (or many) that are out of politicians’ control. Other people who might help determine socio-economic outcomes are civil servants.

In recent years, there’s been more attention to civil servants because people came to realize that they potentially play an important role. And what is the relationship between politicians and their civil servants? The politician can say “Let’s build more schools,” but it is the civil servants who actually implement the construction of those schools. They are going to be responsible for the procurement that leads to the actual construction, and for hiring the teachers who in turn will also become civil servants. Same for hospitals, highways, and so on. Politicians do interact with these civil servants, but we do not know much about how such interaction takes place. For example, do politicians have control over civil servants and, if so, how much? Just think about how challenging it might be to control civil servants: in a given country, there are a few thousand politicians vis-a-vis millions of civil servants.

In a nutshell, Weber says that politicians controlling civil servants is bad, because it leads to patronage. However, under this other perspective, politicians’ control over civil servants improves citizens’ control over politicians, which could lead to better socio-economic outcomes. So who is right?

This is not an easy question to answer. One challenge is data, because data on civil servants are often inaccessible. That is one reason why I focus on Indonesia.^[1] There, like in most other countries, the district mayor has limited power over the local civil servants: salaries depend on tenure and schooling titles, rather than achievements; firing is rare. The mayor might try to influence hiring, but only at the lowest level, because promotions to top civil servant positions can only be internal. Even if you do a lot of patronage in recruitment the people entering the bureaucracy are at the most junior level and thus cannot do anything meaningful for you.^[2] However, I discovered that, in some districts, mayors can influence who, among them, becomes top civil servant. This is one variation I focus on, and I call it “control over the civil service”.

Now, the second factor that should matter politicians’ incentives in office: at one extreme, some politicians may be headed out of office soon (e.g., because of retirement), thus having short time horizon and little/no incentive to perform; at the other extreme, they might have long time horizons conditional on reelection. When it comes to Indonesia, I can use local term limits to compare politicians at the two ends of this spectrum. Hence, I exploit cross-district variation in these institutional features and find that, when politicians control the civil servants and have long time horizons conditional on reelection, they are more likely to demote top civil servants to lower levels. Most of these additional demotions are concentrated among people who were promoted in the previous term by the previous politicians (hence likely loyal to the predecessor). They are also more likely to promote civil servants to top positions, which is also consistent with creating loyalty among the civil servants.

To evaluate whether socio-economic outcomes improve or not, I look at public corruption and households’ consumption. To measure public corruption, I use corruption prosecutions involving public funds and, separately, reports on corruption violations detected based on audits of local governments. To measure households’ consumption, I use multiple rounds of a very large households’ survey that covers most of the country on a yearly basis. In a nutshell, I do find that, in the districts where the mayors have both control and incentives in office, there is both a decrease in local government corruption and an increase in consumption. This is supportive of the theory that some degree of control over the bureaucracy, joint with long time horizons conditional on reelection, is better than bureaucratic insulation. This goes against all the Max Weber type of literature, or at least it goes against an extreme view of the Max Weber perspective on the bureaucracy.

How exactly we go from lower corruption to higher consumption is still unclear. The link between the mayor's control over civil servants (and incentives in office) to households’ consumption seems causal, but the chain of causality is a bit unclear. I used the data on manufacturing firms to see whether the effect was through better firm performance, but I found no evidence of it. It could be that local government officers do a better job in delivering public goods and services. For example, there are many welfare programs that are run by local governments where you either distribute money or food to poor households. If those programs are administered better and there is less corruption there, then you can get better delivery and that can have an impact on the households even without passing through firms. But there is no good data that is consistent across all the surveys so I can’t tell whether that’s true. Perhaps, these are questions for further research.

I do think that my research might have some external validity but I’m not so sure it applies to the Russian setting, because politicians’ incentives in office might depend more on the central government than on the local people. Hence, a similar study could give different results. I actually tried to contact people in Russia to carry out such study (or something related), but have not managed yet: it takes time to build the connections required to do that, you need to build trust, but I hope that one day this will be possible.