STRUCTURE
1. Theoretical framework
In this section we develop a simple descriptive bargaining model of bribery.
We consider, as an illustrative example, a bargaining situation where firms must deal with a number of bureaucratic regulations at a cost of 2rper
regulation.2Obvious examples include compliance with labor safety stan- dards and environmental impact regulations. Firms differ in the number of regulations,nf, they must comply with, based on individual circumstances.
To further simplify, we consider a decentralized model where the firm is engaged in a series of bilateral negotiations with government officials, each of whom may force the firm to comply with the regulatory requirements or pay a bribe to circumvent those requirements.
In each bilateral negotiation, we assume that the regulation may be cost- lessly circumvented by the bureaucrat, so that a surplus of 2ris created by joint agreement to avoid the regulation. The standard Stackelberg bargain- ing solution has the two parties splitting this benefit, so that the bribe will be r. However, in order to reach this agreement, a non-trivial amount of time may be spent negotiating this payment. We assume that, depending on different institutional features, some bureaucratic systems will have an easier time pricing these payments, and therefore the time cost will be lower.
Finally, we allow for the intuition that firms paying numerous bribes will have economies of scale. We thus describe the time spent with bureaucrats, Tf, as g(nf), where 僆(0,1] is a parameter that reflects frictions in the bar- gaining process, with 0 indicating minimum frictions, and g() captures economies of scale in bargaining and is such that g 0 and g 0.
In this highly stylized framework, it follows immediately that the time firms spend in bureaucratic hassle is an increasing function of the level of bribes paid,Bf, since each is a positive function of the number of regula- tions that the firm wishes to circumvent:
Bfnfr Tf g(nf).
The reduced-form relation between bribery and time with bureaucrats is then simply:
(4.1) This example illustrates that, by simply adding negotiating frictions and a firm-specific vulnerability to regulatory hassle, bribes paid are positively correlated with time spent with bureaucrats. This is a straightforward and mechanical result of the model specification where all individual bribes are equal in magnitude (that is, there are no ‘volume discounts’ in bribe pay- ments) and each bribe requires additional time. More interestingly, the pres- ence of the bargaining friction parameter,, indicates that this correlation
Tf g冢Brf冣.
should be weaker under institutions that allow for a relatively efficient nego- tiation process. Hence, our main intuition for the empirical analysis below is that institutional structures that allow for a relatively clear pricing of bribes should be characterized by a weaker association between bribery and time spent with bureaucrats. Bargaining frictions may reflect a number of elements in the bribery negotiation. For example, individuals from similar ethnic or geographic origins may have a common language or frame of ref- erence that facilitates mutual understanding. Repeated interaction between particular business owners and bureaucrats may further smooth this process. In the empirical work below, we shall focus specifically on the fric- tions generated by uncertainty over the amount to be paid in the bribe nego- tiation. In a regression framework, we can capture this effect by running a specification like (4.2) below:
Timef 1*Bribef 2*Uncertaintyf 3*Bribef*Uncertaintyf f, (4.2) where Bribefis a measure of bribe payments by firm f,Timefmeasures the amount of time the firm spends with bureaucratic hassles, and Uncertaintyf reflects uncertainty over the amount of bribe to be paid. Beyond our basic interest in understanding the nature of extralegal relations between bureau- crats and firms, we wish to examine whether a relatively inefficient negoti- ation between the two parties results in slower economic performance. We therefore consider a specification closely paralleling (4.2), where we replace Timeby the firm’s future growth:
Growthf 1*Bribef 2*Uncertaintyf 3*Bribef
*Uncertaintyf f. (4.3)
As a final step, we shall also consider the determinants of country-level uncertainty over the bribe payment by looking at the country characteris- tics that predict average uncertainty by country, that is:
Avg(Uncertaintyf)f(Country Characteristics) c. (4.4) 2. Data
To conduct the empirical exercise, we use data from the World Business Environment Survey (WBS), a firm-level survey carried out in 1999 and 2000 across 61 countries. About 100 firms were interviewed in each country.3 The survey includes basic background information on firms’ characteristics, including number of employees, previous years’ sales, and sector. More importantly, it includes a variety of questions relating to ‘extralegal pay- ments’ to government officials. Among these are the percentage of senior management’s time spent dealing with government officials (TIME) coded
from 1 to 6;4the amount of ‘irregular payments’ paid to government officials, as a fraction of total sales (BRIBE), coded similarly from 1 to 6; and the extent to which firms know in advance how much these ‘irregular payments’
will be (ADPY), coded from 1 to 6, with 6 indicating maximum uncertainty.
In short, these three variables are calibrated so that higher numbers are more undesirable than lower numbers.
Because we are also interested in the effect of the bribe transaction on economic outcomes, we also define a pair of variables relating to the firm’s level of growth. Firms were asked to assess their expected growth rate in sales for the subsequent three years (INCSALES). Because the distribution of future growth projections has very long tails, we consider two transfor- mations of the raw data that place lower weight on outlying observations.
First, we consider an indicator variable denoting whether sales are pro- jected to increase (INCSALESD). Second, to preserve the information on how much sales are projected to change, we consider a log transformation of the following form:
LINCSALESsgn(INCSALES)*log(|INCSALES|)
This variable has the property of being monotonic in INCSALES, but is a much more compressed distribution.
A number of recent contributions have systematized measures of insti- tutional quality across countries. In particular, Djankov et al. (2003) compile a measure of legal formalism across countries reflecting the extent to which the court process is governed by rules rather than discretion in evicting a tenant (FORMAL1) and collecting the payment for a bounced check (FORMAL2). We use a comprehensive measure obtained by simply adding these two measures together (FORMAL).
We also investigate in this context the role of the legal origin of a country.
This variable was introduced in the literature by La Porta et al. (1998) and includes three indicators that classify the legal origin of the company law or commercial code of each country.5 The three classifications of legal heritage are English (common law), French (civil law) and socialist.
Summary statistics for both our firm- and country-level variables are reported in Table 4.1.