The golden goose effect in corruption cases

The Golden Goose

In Economics, an efficiency wage is defined as a wage above the equilibrium price paid to obtain a more efficient job from the employee. For this to be the case, there must be some sort of market failure that prevents the equilibrium from being efficient. One such a scenario occurs when there are some forms of corruption, and was studied first in Becker and Stigler (1974) 1. In this work the authors present a situation in which a principal (e.g., the Government) wants a job to be done and hires an agent to do it (e.g., a public servant), but where imperfect monitoring opens the possibility for the agent to engage in corrupt activities. Becker and Stigler show that by paying and efficiency wage to the agent, corruption may be mitigated: say agents face today an opportunity to cheat and make some extra money, then they have to weight today’s gain against the lost of their job if caught. The lost rent is greater, and then the cheating less attractive, if they are paid more in their actual job (the efficiency wage). Most of the following literature concentrated on this kind of tradeoffs: a cheat today versus a flow of licit incomes in the future.

Recently, Niehaus and Sukhtankar (2013) 2 explore another trade off, namely, the consequences of a cheat today versus a flow of both licit and illicit incomes in the future. If rents from corruption are available in the future, then the agent may not cheat so much in the present for fear of loosing the illicit future rents if caught compared with the situation in which there are no future opportunities for cheating. The authors call this the “golden goose” effect. Not taking into account the effects of future illicit rents may thus weaken the attractiveness of efficiency wages to fight corruption.

In order to measure the “golden goose” effect, these authors use data from the National Rural Employment Guarantee Scheme (NREGS) in India, where a natural experiment took place in 2007 when the state of Orissa changed the pay for daily-wage projects but not for the piece-wage ones. The neighboring, similar state of Andhra Pradesh did not make any change and is suitable to serve as a control for the statistical analysis. The first thing to do is to construct a model in which to study the theoretical effects according to the economic analysis, and then confront them with the actual data. To this end, the relevant characteristics of the NREGS are as follows:

  • Every rural household in India has a right to 100 days of paid work per year within the program, and only needs to apply for it. Within 15 days after application, the worker must be assigned to a project or be given unemployment compensation.
  • Projects are paid in a daily basis or in a piece-rate basis. Both types of projects are similarly lucrative.
  • In every village, the officials in charge of the program keep a digitalized record of attendance of measure output (depending of the kind of projects).
  • The Government reimburses local governments on the basis of these records.
  • Despite some controls, there are two main opportunities for corruption. First, officials may over report the number of working days, and, second, they may pay workers less than the mandatory wage. There are other sources of corruption, like embezzlement from materials, but the study focuses on the first two, as they can be cleanly measured.
  • Workers have no control over the project they are assigned to, and officials have no control over the nature of the project approved for their village.

In the model that captures these features of the NREGS, two major predictions can be shown that are specially useful for empirical contrasting with the observed data: a wage increase in the daily based projects should (i) reduce the theft from piece based projects, and (ii) reduce corruption in villages with more daily-wage projects upcoming compared with the rest of the villages. The rational for (i) is that officials that manage a piece based project do not see their present opportunities changed, but do see that future opportunities for cheating increase, as they could be managing daily-wage projects with higher wages. Prediction (ii) shares a similar rational, as officers in villages with more daily-wages ahead also see more future opportunities for cheating than the others.

There are two sources of data to conduct the empirical analysis. First, the reports filled by the officials, that are posted on a central website, and that include the identity of the workers. Second, the actual pay to the workers, extracted in interviews of a random sampling that includes 1,938 households. The difference between the two sources reveals the size of the corruption. Especial care was taken to make sure that the interviews reflected real data (no incentives to cheat, anonymity and controls, among others).

The analysis suggests that future illicit rents do indeed matter in the way anticipated by the theory. When the daily-wage increased from Rs 55 to Rs 70 on May 1st 2007 in Orissa, theft in this state declined both in absolute terms and relative to the control state of Andhra Pradesh, where wages did not change. Also, the decline was smaller in the villages that had more daily-wage projects waiting in the future, something that is true in all villages, regardless of the kind of project they were currently working on. Although very precise calculations cannot be given with the data, it can be roughly estimated that the “golden goose” effect implies that theft increased by 64% less than it would have had the wage increase been temporary.

The authors find four implications for anticorruption policy:

  1. Future rents matter, a hypothesis that is at the heart of the efficiency wage concept.
  2. Optimal contracts should take both licit and illicit rents into account.
  3. Concerns about the possibility that cracking down one kind of corruption may lead to increases in other kinds should be taken seriously. This hypothesis was first suggested by Yang (2008) 3, and is supported by the data in the present work: “any use of the ‘stick’ that reduces future rent expectations also makes the ‘carrot’ of job security less motivating.”
  4. Policy pilots should be interpreted carefully in weakly institutionalized settings. Simply put, a pilot generates different dynamic incentives than permanent implementation.



  1. Becker, G.S., and Stigler, G.J. 1974. Law Enforcement, Malfeasance, and Compensation of Enforcers. Journal of Legal Studies 3, 1–18.
  2. Niehaus P. (2013). Corruption Dynamics: The Golden Goose Effect, American Economic Journal: Economic Policy, 5 (4) 230-269. DOI:
  3. Yang, D. 2008. Can Enforcement Backfire? Crime Displacement in the Context of Customs Reform in the Philippines. Review of Economics and Statistics 90, 1–14.

3 Trackbacks

Cartografiando la ignorancia #32 | Naukas

[…] Cuando un gobierno contra a agentes para realizar tareas pero no tiene establecidos controles y supervisiones adecuados está dejando huecos para que florezca la corrupción. ¿Podría el establecimiento de sueldos generosos evitar que a los agentes les mereciese la pena corromperse? Esta es el caso que analiza José Luis Ferreira en The golden goose effect in corruption cases. […]

Cartografiando la ignorancia #32 – UMAG – USOMS Magazine

[…] Cuando un gobierno contra a agentes para realizar tareas pero no tiene establecidos controles y supervisiones adecuados está dejando huecos para que florezca la corrupción. ¿Podría el establecimiento de sueldos generosos evitar que a los agentes les mereciese la pena corromperse? Esta es el caso que analiza José Luis Ferreira en The golden goose effect in corruption cases. […]

Ezjakintasunaren kartografia #27 – #29 | Dibulgazioa | Zientzia Kaiera

[…] Hainbat zeregin egiteko gobernu batek kontratatu dituen eragileen lana gainbegiratzeko neurri egokirik ez badu jartzen, ustelkeriak gora egin dezan hutsuneak ahalbidetzen ditu. Kasu honetan, saihestuko litzateke ustelkeria soldata oparoak ordainduz? Hau da José Luis Ferreirak aztertzen duena: The golden goose effect in corruption cases. […]

Leave a reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>