Political parties and public sector reform?


This page in:

Despite years of often fruitful innovation to implement projects through NGOs or community groups or markets, the road to development still passes through the public administration.  And, as development professionals know, the stretch of road in that domain is often impassable and apparently impervious to repair.  Why? 

In “Programmatic Political Parties and Public Sector Reform”, Cesi Cruz and I have found evidence that points to a surprising answer:  countries that lack parties with a reputation for supporting broad public policies, such as those related to the respective roles of the state and markets in economic activity or the relative importance of redistribution and low taxes, are far less likely to adopt public sector reforms.  Programmatic parties appear to be an essential driver of political incentives to pursue public sector reform.

The symptoms of public administration dysfunction are better understood than the disease.  All of the acronymic evaluations, from IEG (Independent Evaluation Group) to QAG (Quality Assurance Group), blame institutional shortcomings for implementation failures across all sectors; they all concur that no sector faces greater challenges to project success than the public sector.  Why?  James Carville, the quotable – and irascible – political consultant from New Orleans, would surely say (even when speaking to very smart people):  “It’s the politics, stupid.” 
The most sophisticated research examining the intersection of politics and public administration focuses on successful democracies.  It asks: how do legislatures and coalition partners shape the institutions of public administration to ensure that it implements the policies they have approved?   The research takes for granted that they care, a reasonable assumption in advanced democracies.  In many of the countries in which development professionals operate, it is not.

More common is the experience I had when I recently spoke individually with more than ten MPs in a West African country.  We discussed legislative oversight of the budget, but never for very long:  all acknowledged with apathy or resignation that such oversight scarcely existed.  Donors had funded a secretariat with a smart and engaged staff to enable the legislators to better question executive proposals and implementation.  The staff reported little interest in their efforts and presented ample evidence to demonstrate the indifference of legislators to oversight of the executive.

Such indifference raises a puzzle that both the academic literature and the strategies of donors have largely ignored:  politicians who are disinterested in the administrative apparatus through which legislation is implemented.  To explain this puzzle, Cesi and I made the following observations.

First, politicians whose futures depend on the fulfillment of broad programmatic promises are more likely to need the services of a well-functioning public administration than politicians who, instead, depend on patronage appointments and the manipulation of regulatory decision making in the service of cronies.  Second, brave legislators who challenge the executive branch on their own, without the support of their colleagues, can rarely overcome the resistance of the executive and the executive’s ability to play “divide and conquer”.  In contrast, legislators who are collectively organized under the umbrella of a programmatic party not only have a stronger incentive to monitor executive behavior, but are also better to resist executive intimidation.

The evidence supports these observations.  We took advantage of data on the performance of 511 Bank public sector reform loans that the Internal Evaluation Group assembled for the recent IEG review of World Bank public sector activities.  These loans are fine proxies for public sector reform success.  The start date of the loan tells us when the process of public sector reform began.  A variety of IEG assessments tell us how the reform effort ended.  They reflect the difficulties of public sector reform:  1.4 percent of the loans merited a high rating on institutional development; 39 percent had a lesser, but substantial impact on institutional development; 48 percent had only modest effects.  Almost ten times as many loans, 12 percent, had a negligible effect as received the highest rating. 

The Database of Political Institutions has more than 100 objective variables on political systems (from the number of years the executive has been in office to the competitiveness of elections to the number of political checks and balances) for more than150 countries over the period 1975-2009.  It provided us with information on whether parties were programmatic.  Though a rough indicator, it matches up nicely with several more detailed investigations of party characteristics that scholars have undertaken at different points in time for smaller groups of countries.  The programmatic party variable goes from zero to one, averaging .51 over the sample. 

The figure below illustrates the substantial effects of programmatic parties on public sector reform.  When there are no programmatic parties, and controlling for numerous other country characteristics, the probability of an IEG project succeeding is only 20 percent (and, with 95 percent confidence, between 6 and 33 percent); when all parties are programmatic, the probability is 59 percent (or, with 95 percent confidence, between 46 and 73 percent). 

Is this simply a coincidence?  If so, the coincidence persists under numerous estimation approaches.  We controlled for other political explanations for reform – the extent of democracy, the nature of electoral institutions, political instability; for reliance on natural resources, country income, a subjective measure of bureaucratic quality at the outset of the loan (as measured by International Country Risk Guide); and for country size, the size of the loan, or its duration.  We excluded reverse causality as an explanation (that IEG gives higher ratings to countries with programmatic parties); and unobserved phenomena that might explain both the existence of programmatic parties and successful reform.  Coincidence is unlikely.

In discussing the timing of reform and of donor collaboration with client countries, practitioners and donors emphasize that one should take advantage of windows of opportunity. Those windows are usually considered open when influential politicians demonstrate enthusiasm for reform. Our analysis holds enthusiasm constant: all loans generated enough enthusiasm to persuade a few top ministers, including the one in charge of the relevant sector, to approve the loan. The individual commitment of a few turns out to be insufficient, however. In addition, and critically, reform succeeds when a larger group of politicians benefits collectively from it – that is, when politicians are gathered under the umbrella of a programmatic political party.



Philip Keefer

Principal Advisor, Institutions for Development at Inter-American Development Bank

November 11, 2010


As with all great insights, this one looks so much more obvious after somebody has made the point clearly, as you and your co-author have.

Some reactions to set the ball rolling on this issue -

A programmatic political platform is itself an indication that a certain set of collective action problems associated with public policy and public good provision is partly addressed. Therefore one might be a little more optimistic that such a government will attempt to strengthen public management institutions and practices to achieve its platform objectives.

One complication to this story in aid dependent countries - can we really conclude objectively whether countries have political parties with such programs? We need to be conscious of the problem donors create by attempting to influence the definition of programmatic policy through well-meaning attempts to promote development policy (the poverty reduction strategy, MDG commitments etc.) or other variants of an externally prompted policy program. So, we might observe a policy framework that looks programmatic (and opposition parties might also mirror this in some way to be aid-competitive) but is largely the result of donor prompting and in fact there is often no organic political process from which it is derived.

So we might have some trouble applying your insight to make useful judgments about the prospects for PSR in donor dependent environments because it will be difficult to tell the real nature of political programs from what is publicly advocated by political parties.

A broader question then is whether donor influence retards the natural development of political party platforms and whether isomorphic mimicry is not also a problem at this high policy level.

Phil Keefer
November 14, 2010

Zubair Bhatti sent the following great article that he published in Dawn before joining the Bank; it describes the consequences for public sector performance, at the local level, of party-less political competition.


Phil Keefer
November 12, 2010


Great comment (and great title). To make cross-country comparisons, we have to rely on relatively crude data on parties. This is changing: Herbert Kitschelt from Duke University spent a morning at the Bank two weeks ago presenting a new, comprehensive database of essentially every political party in about 85 countries (the World Bank Research Support Budget was an important source of financing for the effort).

Looking at a specific country, though, we can have access to much more information (e.g., from Afrobarometer surveys) that allows us to establish two conditions:

Are politicians from a party bound together in a common enterprise? The answer is more likely to be yes if voters care about party affiliation and not just about the party leader; probably no if candidate-financed gift-giving is the secret to political success and party switching is rampant.

Is that common enterprise the pursuit of programmatic goals? The answer is more likely to be yes if candidates are selected on the basis of adherence to the goals and if voters who share those goals are the ones supporting the party. Otherwise, the common enterprise is more likely to be the feeding of the party machine.

If facts on the ground tell us that the answer to both questions is no, broad public sector reform (actually, any broad reform to improve public good provision) is least likely to be sustainable. More focused reforms (on sector and region) are more likely to succeed. If both are true, we can easily discard the possibility of isomorphic mimicry and feel more confident in pursuing public sector reform.

The links with foreign aid need to be better understood. Lots of research suggests that in some settings, foreign aid is viewed as a rent to be extracted. High-rent environments (whether aid or natural resources) are not great for building organizations meant to pursue collective interests. On the other hand, foreign aid has also contributed to organization-building, whether through bilateral assistance to parties, or general donor assistance to large NGOs. We don't know, however, whether such efforts have transformed these organizations into vehicles for pursuing broad citizen interests. We need to find out.

jacob thompson
March 05, 2011

I had no idea that political parties played such an important role in these countries politics. I am aware that politics happens but not to this extent

Mascha Rauschenbach
December 08, 2011

Dear All,

I am awaiting these new data with information on whether parties are programmatic or clientelistic with impatience and have a few questions concerning this. Does somebody know already how many parties per country will be included and when the data will be released? Will they be included in the next release of the Database of Political Institutions and when will the newest update of the DPI be publicly available?
In my doctoral dissertation on parties in Africa I am currently measuring party strategy (programmatic or clientelistic) with information on party ideology. I follow Phil Keefer's approach to measure a party as clientelistic if it is coded as having no ideology and as programmatic if coded as being left, right or center. But I would like to know a little bit more in detail how this ideology variable was coded, than the information the DPI codebook provides. I would be grateful for any help!