Bribery in Indonesia: some evidence from micro-level data

Abstract

This paper outlines and tests a model in which firms seek to reduce the cost of taxes and regulatory compliance by offering bribes to government officials. It finds that firms' profitability (scaled by production costs) largely determines both the amounts paid and the time spent negotiating bribes with officials. Competition between arms of the bureaucracy for bribe income seems to be a result of decentralisation, but the analysis suggests that this competition would lead to a spreading of bribes among a larger number of officials rather than to a significant increase in their total amount. Local governments may be able to raise more revenue by reducing the number of taxes and regulations and using part of the increased revenue to raise the salaries of officials, while devoting more effort to restraining corrupt behaviour. But progress may be blocked by central government tax officials increasing their demands for bribes.

Acknowledgements

I am grateful to Andrew MacIntyre and two anonymous referees, and to seminar participants at the Australian National University, for many helpful comments. All remaining errors are mine. I would like to thank the team members of the University of Indonesia's Special Survey on Governance, in particular, Thia Jasmina, Ari Damayanti, Isfandiari Jafar, Isfandiarni, Suryadi and Eugenia Mardanugraha.

Notes

By mid 2004 the number of districts was closer to 400.

Firms that did not report paying bribes may in fact have paid them, of course, and this has important implications for the analysis. These are discussed in detail later.

These six districts, chosen to reflect the geographic spread of firms in Indonesia, were Bandung, Semarang, Malang, Palembang, Pontianak and Makasar.

The mean values shown here differ from those in table 1 in being unweighted: a district with 10 firms counts equally with another with 30.

Interestingly, the five municipalities of Jakarta are among very few that claim that PAD revenue is high enough to cover both routine and development expenditures. Motor vehicle tax is the most significant contributor to PAD.

Street lighting and hotels and restaurants were favourite targets. The street lighting tax is collected by the state electricity company (PLN) on behalf of the local government as a surcharge on its customers.

Any other tax levied on these firms would have the same impact, of course.

In effect, firms have to obtain permission to make a ‘disturbance’, such as noise, traffic congestion and waste discharge, before the business can be operated.

A well-known joke has it that, when asked why he robbed banks, a thief replied: ‘because that's where the money is’. By analogy, the predator official can be expected to focus his attention on the firms he or she believes to be most profitable.

By analogy, if a rational buyer in a food market thinks that the seller's weighing scales have a downward bias, she or he will offer a correspondingly lower price per kilogram of fruit.

The United States Foreign Corrupt Practice Act (1997); the OECD Convention on Combating Bribery of Foreign Public Officials in International Business Transactions (1999).

Tax offices in the districts and municipalities are not part of local government, but rather of the central government machinery for collecting central taxes such as corporate income tax. This has important implications for anti-corruption efforts at the district level, as we shall see later.

At the beginning of the survey, respondents were asked the value of annual sales, but the response rate was very low. It was therefore decided subsequently to ask respondents only to place their firms within one of these four size categories.

We performed several experiments in which the share of districts’ output of other natural resource products in GDP—specifically, timber and minerals—was included in the regression, but these variables had very low statistical significance and so were dropped from the analysis.

For example, the variable used to measure bribe efficacy in the empirical model should correlate positively with alternative measures of bribe uncertainty, such as the firm's ability to predict the bribe level.

For example,time spent with officials to smooth business operations should correlate positively with the degree of difficulty in getting officials to process business permit applications.

Recognising the secretive nature of bribery, we experimented with different question sequencing in the pre-test questionnaire. On the basis of these experiments, the question on bribes paid (as a percentage of production costs) was asked roughly half-way through the interview.

For the Heckman procedure, only the two best performing specifications (models I and II) are presented for comparison with OLS and Tobit.

Lambda is a new, constructed variable containing an inverse Mills ratio obtained from the Probit regression, and calculated as the ratio of the posterior density function to the cumulative density function of the standard normal distribution (Heckman 1979).

The coefficient of bribe efficacy in the Probit equation (table 5) is negative and significant at the 5% level.

This is weakly supported by a correlation coefficient between tax payments and time spent of 0.16.

The relative weakness of the correlation between the regulatory burden and the bribe rate is not caused by multi-colinearity with time spent, since the coefficient correlation is only 0.12. Furthermore, the coefficient of time spent is statistically still strong.

The weakness of this result is not a multi-colinearity problem with either the regulatory burden or the time spent variable, since their correlation coefficients are only 0.026 and 0.016 respectively.

There is a strong correlation between firm age and firm size, so age is excluded from the model specification that includes size.