Weber’s Bureaucracy & Dominican Republic

As said by Johnson & Libecap about US, “Cynicism about the federal bureaucracy is widespread and the general public views federal employees as aloof, uncaring bureaucrats who are unresponsive to their requests”. So, it must be unravelling as well when it comes to a clientelist state like the Dominican Republic (DR).

DR is considered to be a highly clientelist state (38% public servants worked in political campaigns) ranked 3rd out of 88 countries (Kitschelt, 2014) with political masters demanding and receiving clientelistic services from the politically “appointed” bureaucrats, corruption was found at high levels, scoring 0.81 in the V-Dem Index [CaseBook 1]. All this brings DR in the top bracket of highly corrupt countries along with high historical discretion/patronage state in the matters of selection, recruitment, pay, perks and dismissal. It stood out on these with greatest range and depth in an expert survey of 22 countries (Kopecky, 2016). Some merit- based selection  started in 2004 [2% of total vacancies in about 65 state institutions filled based on merit] and about 7% of total employees were incorporated into public service since 1995. Largely, DR provided no ‘tenures’ and hence no job stability (political patronage decided whether they remain in their jobs).

So why does it happen that some states flourish while others become a case study in public maladministration, bureaucratic loyalty to political bosses, financial bungling, low motivation and low honesty & integrity of bureaucrats. The authors were thus motivated to test the theory propounded by Weber, more than a hundred years ago, which links well-defined bureaucratic structures to higher motivation, lower corruption and low clientelism. Oliveros and Schuster’s *CaseBook 1* is thus an effort to test Weber’s theory of causes for bureaucrats’ behaviour, in DR’s context.

The authors applied “Conjoint Survey Experiment” in a government institution [General Audit Office] where unequal ‘treatment’ of merit & tenure was the norm. They developed 6 hypotheses to test, based on 2 variables of merit & Tenure and 3 dependent variables of characteristics i.e. motivation, corruption & clientelism with 6 attributes (i) year of appointment, (ii) form of recruitment, (iii) administrative career (tenure), (iv) education, (v) position, and (vi) gender. Out of 6 hypotheses produced, 4 were proved and 2 disproved. For example, Hypothesis 1 (H1) was:  “Public employees recruited through examinations will be less likely to engage in corrupt behaviour” [CaseBook1]. The methodology followed was that each respondent was emailed with five sets of hypothetical pairs of bureaucrats’ profiles with randomly varying degrees of the above characteristics. These sets were randomly varied on above six attributes and the order of questions also randomised to rule out primacy effects but kept fixed across pairings (to reduce complexity). Questions were asked e.g. (i) “Which of the two would you trust to administer the funds of a project transparently?” [CaseBook 1] as an indirect measure of corruption and similar indirect ones were framed to measure clientelism and motivational levels. The design helped authors to isolate the effects of bureaucratic structures on all the studied three characteristics by utilising ‘Linear Probability Model’. LPM is a type of binary regression with only two dependent variable states and probability of the two is then modelled. The specialty of the method was that it did not ask specifically what the personal preferences of respondents were, or the reasons for preferring one profile over another to remove social desirability bias. 

Overall, the Weberian ‘merit based selection’ was found to lead to higher motivation, lower corruption and lower clientelist services and ‘tenure’ produced low clientelism whereas low corruption and high motivation were either found to be spuriously related during robustness checks or without any significant effect. Statistically, a 10% average marginal effect was observed in case of merit based employees being trusted, similarly a greater trust was found to be put in educated, female, and technical-professional public servants towards lower corruption. Respondents also put a -12% reduction in participating in electoral campaign (clientelism) when the public servant was recruited through merit. Similarly, tenure (job stability) produced a negative effect of -8% on political clientelism i.e. respondents said that those with job stability were hard to convince to participate in electoral campaigns. It was also indicated in the responses that it was difficult for female employees to be convinced to attend political rallies (low clientelism). Another significant result was that it was 7% more likely that an employee selected on merit would be more easily motivated to work hard. A larger ‘motivational’ effect (16%) on ‘untenured’ versus ‘tenured’ employees was spuriously found as it meant that instead of ‘tenure’, some unobservable characteristics might be at play, so tenure actually did not have any effect on motivation. Also, during the robustness checks (for confounding), the tenure effect on corruption was not found to be robust though initial screening had given a large causal effect, meaning thereby respondents might have chosen this behaviour in tenured employees based on some unobservable characteristics (perhaps closer relationships with boss or some other skills).

Critique of Study: Conjoint allows to identify, measure and compare the independent effects of various characteristics in a single experiment (Hainmueller, Hopkins, & Yamamoto, 2014), however, heteroskedasticity is a part of this method and needs to be corrected (Murray James M) since by definition probabilities are bound between 0 & 1, & linear equations having no bounds. Yet, the results/discussions do not talk about it. It was better than Observational studies in that ‘randomisation of attributes’ removed omitted variables and reverse causality biases. Other biases such as social desirability bias was taken care of and allowed to estimate the effects of different attributes simultaneously, which was more realistic in the sense that it allowed trade-offs between preferences for different characteristics. However, the outcomes of the study were perception based rather than ‘actually measuring’ the bureaucratic behaviour.

Following biases were also found in the study: (i) sample representativeness was limited to  gender and age and based on only one institution where more experienced & professionally educated bureaucrats might me working, so making sampling none too representative, however, sub groups were then formed for robustness and out of 36 so formed, 35 subgroups did show validity, even though a “totally representative survey” yielding different results can’t be negated (ii) respondents could be answering strategically about government’s reform program more favourably, and/or they might use their prior knowledge-based responses about merit & tenure rather than on their actual field-based experience, (iii) the ‘halo effect’ where the respondents see fellow colleagues favourably when they share their own perceptions and characteristics. (iv) The authors claims to have done robustness checks and established their results’ validity based on sub-groups formation, though the authors have sparsely used effective visualisations, except Point Estimates, in main text and the quoted Appendix (having tables, charts & graphs) link did not work [Page 24: Supplemental Material:Online Appendix http://journals.sagepub.com/doi/suppl/10.1177/0010414017710268]. (v) Finally, the current experiment does not talk much about the impact of ‘Information Equivalent (IE)’ (Dafoe, Zhang, Caughey, 2018), just a passing comment was made.

Following are my recommendations:

1. The study results are robust with good internal validity and established causal links described above, the policy seeks to enhance ‘efficiency’ of the bureaucratic system. And since both merit and tenure is very thinly spread across DR’s institutions, a phased scaling up of the policy is recommended.
   
2. A mid term impact evaluation must be carried out, let us say, after 3-5 years of first scaling up. This should validate continued applicability of the policy. Subject to confirmation of similar results, further scaling up should continue to cover the vast bureaucracy of DR. 
   
3. To mitigate further the concerns of reverse causality biases and omitted variables due to it being a single country study which might affect its external validity, a study similar to Meyer-Sahling, Mikkelsen, Schuster (2020) on a  cross-country design must be followed to see if the results are transferable.
   
4. To commission another study after a similar period of 3-5 years to gauge the impact on prospective candidates as to how people get motivated to join bureaucracy once the merit based selection & tenures effects are realised across the country. This will help to understand what really works & how, which is a missing cog in the wheel here.

The Strengths/Weakness/Challenges:

As per Principal Agent theory, the Bureaucrats are Agents appointed by their politician Principals. And hence there is a chance that they may indulge in shirking and slippage (Agency loss) and thereby it is the responsibility of the Principal to create such an environment and contracts/agreements that these losses are minimised. This can be achieved by appointing people based on solid criteria of merit and giving them job stability where there is no fear of being fired unilaterally, which will help them to concentrate on their official assignments. Hence, this policy is robust on this ground. Moreover shifting their loyalty from politicians to the state will reduce Moral hazard too as they will be less and less inclined to fund politicians to stay in their jobs. 

Application of this policy will also give bargaining power to bureaucrats and will help them to a certain extent in raising the issues of state’s interest when politicians themselves indulge in corrupt activities and similarly expect bureaucrats to do the same on their behalf; instead, now bureaucrats can create an ‘interest group’ to gain bargaining power with politicians, though too much bargaining power could prove to be a two-edge sword as in the case of Presidential system of US (Johnson & Libecap,1994).

Bureaucrats recruited through merit examinations and given tenure indulge less in corruption and, not being indebted to their political bosses, are then also able to provide good governance which further enhances delivery of public services. A further strong point which came out in the study is that females indulge in less corruption/clientelism and so gender-based governance research opens up from this study and paves way for women’s representation in bureaucracy.

Some challenges, though, could be faced during the scaling up of policy. First and foremost since it is an elite-controlled agenda,  it is difficult to bring this policy. Notwithstanding the top leadership thinking, large scale political resistance will come in the way since most politicians will be loath to lose power through these ‘institutional reforms’ that make clientelism difficult, potentially undermining their electoral prospects too.

Moreover, since DR is a clientelist state, to change the culture in the beginning will be quite a herculean task for bureaucrats themselves as there will be a lot of resistance from within the ‘clientelist bureaucracy’ too. Hence during the policy cycle, an impact evaluation will be required to know the efficiency and the areas of improvement in its implementation.

The Principal Agent approach to corruption as governed by the corruption equation C = D+M-A as proposed by Robert Klitgaard (2008), means that discretion and monopoly should be reduced and accountability increased. Here we are giving the bureaucrats job stability i.e they could not be easily removed from services which should mean that they are more free to carry out their mandated duties rather than indulging in political services but at the same time now politicians will be monitoring them closely as the corruption  and clientelist linkage between them is weakened or even broken, thus indirectly increasing accountability. So, overall, this will help in reducing corruption from this perspective too.

However, we have to be careful for there are areas of concern too: the Weberian bureaucratic structures, as per the study, does not reduce corruption and does not increase morale by giving tenures  and incorporating the bureaucrats into public service. The empirical literature in this field is scarce. However, a study on the American case where the bureaucracy enjoys all the benefits of bureaucratic structures shows that great concerns remain about bureaucracy there as well (Johnson & Libecap,1994). Hence, again, we have to go for a mid-term impact evaluation after a partial scaling up of policy.

References:

1. Bansak, Hainmueller, Hopkins, Yamamoto (Sept, 2019).  “Conjoint Survey Experiments” For Druckman, James N., and Donald P. Green, eds. Cambridge University Press. 
   
2. Christian Schuster (January 2015). Institutional Incentives for Professionalizing Patronage States [Dissertation of Christian Schuster].
3. CaseBook1: Oliveros & Schuster (2017). Merit, tenure, and bureaucratic behaviour: Evidence from a conjoint experiment in the Dominican Republic
   
4. Chapter 7: OECD. Available at https://www.oecd-ilibrary.org/sites/270b68d2-en/index.html?itemId=/content/component/270b68d2-en
   
5. Colonnelli, Prem & Teso. Patronage and Selection in Public Sector Organizations.
   
6. Dafoe, Zhang, Caughey (August 2018). Information Equivalence in Survey Experiments. Published online by Cambridge University
   
7. Guillermo Toral (2021). The benefits of patronage: How political appointments can enhance bureaucratic accountability and effectiveness.
   
8. Gimpelson & Treisman Fiscal Games and Public Employment: A Theory with Evidence from Russia
   
9. Hainmueller, Hangartner, Yamamoto (2015). Validating vignette and conjoint survey experiments against real-world behaviour
   
10. Harris, Meyer-Sahling, Mikkelsen, Schuster.  Activating the “Big Man”: Social Status, Patronage. Networks, and Pro-Social Behaviour in African Bureaucracies
    
11. Johnson & Libecap (1994). The Federal Civil Service System and The Problem of Bureaucracy, The Economics and Politics of Institutional Change.
    
12. Kitschelt (2011). Clientelistic Linkage Strategies. A Descriptive Exploration
13. Klitgaard, R (Jan, 2008). A Holistic Approach to the Fight against Corruption Bali, Indonesia.
    
14. Kopecky (2016). Party patronage in contemporary democracies: Results from an expert survey in 22 countries from five regions.
    
15. Krause, Lewis & Douglas. Political Appointments, Civil Service Systems, and Bureaucratic Competence: Organisational Balancing and Executive Branch Revenue Forecasts in the American States 
    
16. Kurer (August 1991). Clientelism, corruption, and the allocation of resources. Queensland University of Technology, Australia.
    
17. Meyer-Sahling, Mikkelsen, Schuster (2020). Merit, Recruitment, Tenure Protections and Public Service Motivation: Evidence from a Conjoint Experiment with 7300 Public Servants in Latin America, Africa and Eastern Europe.
    
18. Muñoz & Prem. Managers’ Productivity and Recruitment in the Public Sector.
    
19. Murray James M. University of Wisconsin – La Crosse. https://murraylax.org/rtutorials/linearprob.html#introduction
    
20. Robinson & Verdier (2013). The Political Economy of Clientelism
    
21. Singer, M ( Buying Voters with Dirty Money: The Relationship between Clientelism and Corruption.
    
22. UNODC (Nov 2009). Quantitative approaches to assess and describe corruption and the role of UNODC in supporting countries in performing such assessments Background paper prepared by the Secretariat.