November 2016|bsi|

Development practitioners are rightly concerned with depoliticising clientelist civil services. In other words that the recruitment of officials be based on merit rather than political loyalty for all but the top echelons of the bureaucracy. A number of statistical studies have linked depoliticised bureaucracies to positive development outcomes, including increased economic growth, lower levels of corruptionand poverty reduction.

At the same time, the track record of donors assisting such depoliticisation has been anything but stellar. The World Bank, for example, noted in 2008 that the $422m it lent each year for civil service and administrative reform between 2000 and 2006 had no measurable impact. The prime obstacles to change were found to be political rather than technical in nature.

Awareness of the potential impact of reforming clientelist civil services, along with an understanding of the political difficulties of doing so, has led scholars and practitioners alike to take a growing interest in the politics underlying such reforms. A range of recent studies have explored the conditions under which politicians may be willing to depoliticise civil services, and why they tend to depoliticise some parts of the state – so-termed ‘islands of excellence’ – ahead of others.

However, it is crucial to recognize that politicians not only depoliticise some parts of the state ahead of others, but also some personnel decisions ahead of others. For example, in several Latin American countries civil servants are recruited on the basis of clientelism yet are protected from dismissal by tenure regulations. Thus, political considerations may still determine the hiring and promotion of bureaucrats, but they can no longer be fired at will.

But why are politicians willing to give up control over some parts of the bureaucratic career ahead of others? And does depoliticising one part of the bureaucratic career encourage or hinder the depoliticisation of other parts?

Exploring the interplay between regulations protecting civil servants against dismissal (tenure) and meritocratic recruitment and promotion (merit) as two key parts of the bureaucratic career may shed some initial light on these thus far neglected questions.

Contrary to conventional wisdom concerning merit and tenure, I will argue that merit tends to follow tenure when depoliticising civil services. Put simply: politicians (the patrons) may grant job protections to bureaucrats (the clients) when uncertain about their re-election. When unable to reverse these tenure protections, future incumbents are less able to compete electorally based on clientelism and are therefore more likely to introduce merit.

Let me explain this sequential argument in a bit more depth (feedback on the corresponding conference paper is greatly appreciated). In bureaucracies where all personnel decisions are based on political discretion (i.e. in which neither tenure nor merit have been introduced), uncertainty about re-election may incentivise incumbents to safeguard the job stability of their appointees. Consequently, they seek to hijack capacity-oriented civil service reform programmes – often supported by donors – to introduce tenure protections. This is frequently lamented by donors as the “merit trap”, not least because, according to recent statistical evidence, while merit-based personnel decisions yield the aforementioned positive development outcomes, tenure protections have, on average, no effect.

Subsequent administrations that inherit a bureaucracy staffed with their predecessor’s appointees will want to reverse these tenure protections to maximize their clientelist powers – including the power to fire and replace the entire bureaucratic stock. But when veto players thwart such reversals, the incentives of incumbents may shift. Being unable to substitute the inherited civil service with their own appointees, incumbents are deprived of a large amount of their clientelist powers. As such, they will be less able to compete electorally based on private goods. Consequently, they may introduce merit to at least enhance the public goods they may derive from their de facto civil services. In short (and probabilistically), merit follows tenure. So the much deplored “merit trap” may actually constrain clientelism for subsequent administrations and thus shift their incentives towards introducing merit.

Analysis of merit and tenure data1 from the University of Gothenburg’s Quality of Government Expert Survey 2008-10 seems to bear this out. Merit is five times more frequent when tenure is also in place2; tenure is only 1.8 times more frequent when merit is also in place. Tenure is 2.2 times more frequent without merit (relative to tenure with merit) as compared to merit without tenure (relative to merit with tenure). A tentative3 statistical test provides further evidence for the finding that merit follows tenure. When controlling for the key determinants of merit put forward in other studies (socio-economic factors such as income, education, ethnic fractionalization and urbanization; political-institutional factors such as legislative competition, veto players, political time horizons, government revenue and democratization; and external factors such as aid and economic integration) tenure remains a statistically significant determinant of merit.

It’s important to note that this is a preliminary, probabilistic rather than deterministic, finding. Merit far from always follows tenure. Longer political time horizons, higher per capita incomes, greater press freedoms and more veto players also impinge upon the introduction of merit in the statistical test presented above. They will similarly play a role in shifting incentives for politicians in favour of forsaking clientelist control over bureaucratic careers.

Perhaps more importantly, looking at the contribution of tenure to the introduction of merit allows for broader conclusions about sequences and unintended consequences of reforms to developing country bureaucracies. It puts a premium on considering not only the immediate technical results of support to state reform efforts but also their effects on political incentives for future reforms. Sometimes an unintended and much-deplored initial result – in this case the “merit trap” of tenure protections – may, further down the road, unintentionally shift political incentives towards the reform desired in the first place.


1. Merit may be usefully captured by responses (on a 1-7 scale) to the survey question: “When recruiting public sector employees, the skills and merits of the applicants decide who gets the job?” Tenure protections in turn may be usefully approximated by responses to the question: “Once one is recruited as a public sector employee, one stays a public sector employee for the rest of one’s career?”
2. Based on dummies for merit and tenure set respectively to 1 whenever survey responses exceed the mean value of the response scale (=4).
3. The test is tentative not least as the lack of panel data on tenure and merit precludes a more conclusive statistical analysis of their sequential relationship.

$permalink = the_permalink();