Is the Internet an Effective Mechanism for Reducing Corruption Experience? Evidence from a Cross-Section of Countries

Transcription

1 Is the Internet an Effective Mechanism for Reducing Corruption Experience? Evidence from a Cross-Section of Countries Jamie Bologna College of Business and Economics West Virginia University Morgantown, WV em : Jamie.Bologna@mail.wvu.edu Abstract: This paper develops an indicator of Internet awareness of corruption as described in Goel et al. (2012) to see how this impacts both corruption perceptions and corruption experience. The results confirm the finding of Goel et al. (2012) that corruption perceptions are highly influenced by Internet awareness. However, the effect Internet awareness has on corruption experience is unclear. This paper finds that Internet awareness decreases the frequency of corruption experience of households, while it increases the frequency of corruption experience in firms. Overall, the results suggest that the effect Internet awareness has on corruption is highly sensitive to the corruption measures used and the time the Internet data is constructed. Keywords: Corruption perceptions, corruption experience, Internet, press JEL Codes: D73, K40

2 I. Introduction The negative relationship between corruption perceptions and economic growth is fairly robust (Treisman, 2007). The literature suggests using media to hinder corruption by increasing the likelihood that corrupt acts will be discovered and therefore enabling voters to punish corrupt officials. In general, there have been several studies showing that a free press and media attention tends to reduce corruption perceptions (Djankov, 2001; Brunetti and Weder, 2003; Chowdury, 2004). However, these studies are mostly analyzing media s impact on corruption perceptions, not experience. As Donchev and Ujhelyi (2013) explain, individuals may base their perceptions of corruption on events publicized in the media and not experience, increasing their awareness. In general, there is found to be a relatively weak relationship between corruption experience and corruption perceptions (Weber Abramo, 2008; Mocan, 2008; Donchev and Ujhelyi, 2013). In addition, Bologna (2013) finds that corruption perceptions and experience have different impacts on growth. Therefore, an important task would be to determine if media coverage is only impacting corruption perceptions, corruption experience, or both. Goel, Nelson, and Naretta (2012) develop a measure of Internet awareness to show that this awareness is negatively correlated with corruption perceptions as well as a single measure of corruption experience in firms. The goal of this paper is to expand upon the Goel et al. (2012) analysis by including additional experience measures. The perceptions indices represent a mixture of corruption in both firms and households, and have been argued to measure petty corruption rather than grand corruption (Kenny, 2009). Thus, including measures of corruption that represent corruption experience in households as well as experience severity would be important when comparing the effect of Internet awareness on these measures. In addition, since

3 the content of the Internet changes on a daily basis it is important to check the robustness of the findings in Goel et al. (2012) with more years of data. II. Internet Awareness and Corruption Data This paper develops a measure of Internet corruption awareness and other controls that are likely to impact these corruption perceptions and corruption experience measures as described in Goel et al. (2012). 1 This measure is comprised of Internet searches for country name with corruption and bribery using Google and Yahoo; e.g. for Albania I searched Albania corruption bribery in both search engines and recorded the number of hits for each. 2 The number of hits is then averaged and divided by the population in that country. The corruption data comes from three sources: (1) World Bank (WB), (2) Transparency International (TI), and (3) United Nations Interregional Crime and Justice Research Institute (UNICRI). I follow Goel et al. (2012) and use TI s corruption perception index and WB s control of corruption index as measures of corruption perceptions. These indices are inverted such that a higher value means more corruption. In addition, this paper uses the WB s World Business Environment Survey (WBES) as in Goel et al. (2012) to measure corruption experience. The WBES, conducted in the , surveyed over 10,000 firms in 80 countries on both their experiences and perceptions of corruption, among other things. Each firm was read the following sentence: It is common for firms in my line of business to pay some irregular "additional" payment to get things done." The firms were then asked if this was true, on a scale of 1 (always) to 6 (never). This index is inverted so that a higher number means more corruption. 1 See Goel et al. (2012) for details on construction of the Internet measure, control variables, and corruption perception measures. Goel et al. (2012) use measures of institutional quality as controls from the International Country Risk Guide; due to data availability, this paper uses equivalent WB governance indicators. 2 The Internet search for this paper was conducted on October 31, 2013.

4 The WBES also asks each firm On average, what percent of revenues do firms like yours typically pay per annum in unofficial payments? The answers are recorded on a scale from 1 (0%) to 7 (over 25%). 3 This measure is capturing the severity of corruption and is not included in the Goel et al. (2012) analysis. This paper also includes two measures of household corruption experience. The first is from UNICRI s international survey of households across countries in 1996 and 2000, asking individuals if they were asked or expected to pay a bribe in the past year. The second household experience measure comes from TI s Global Corruption Barometer (GCB) survey, asking individuals if they paid a bribe in the past year. This paper uses these surveys as estimates of frequency of corruption experience by recording the percentage of households that responded yes to this question in each country. III. Empirical Specification and Results This paper estimates the following linear relationship: (1) _ i = countries j = CPI, CC, APAY, BRIBE_WBES, BRIBE_ICVS, BRIBE_GCB k = GDP_PC, G_GDP m = FREEDOM, VOICE, EFFECTIVE, RULE. Thus, the dependent variable is one of the six different measures of corruption where a higher value means more corruption. In each specification INTERNET_HITS, GDP_PC, G_GDP, and FREEDOM are included to control for Internet awareness, economic prosperity, size of government, and general institutional quality. We refer to these four controls as the full set of 3 This paper follows Batra et al. (2003), by using the midpoint of each category to calculate the total percent of sales firms pay in bribes.

5 controls. In addition the responsiveness of government to citizen demands (VOICE), quality of government (EFFECTIVE), and rule of law (RULE) are separately included to estimate their effect on corruption. In all specifications that aren t included in the Goel et al. (2012) analysis, controls are included for the year that corresponds to the corruption measure. Brief data descriptions and sources are given in Table 1. Each results table of the appendix presents on the coefficient on INTERNET_HITS for brevity and includes four specifications. The first include the full set of controls only and the last three include the full set of controls in addition to VOICE, EFFECTIVE, and RULE in specifications 2, 3, and 4, respectively. Table 2 reproduces the main results in Goel et al. (2012) and all of the coefficients have same sign for the corruption perceptions measures (panel A and B). However, INTERNET_HITS is now statistically significant at the one percent level in all specifications and the opposite sign for corruption experience (panel C). Table 3 presents the results for the other three measures of corruption experience. Only two of the specifications have statistically significant coefficients; BRIBE_UNICRI is statistically significant in specifications three and four. In addition, the signs in panel A and B are negative, while the signs in panel C are positive. Lastly, Table 4 presents the updated results of the corruption perception measures using the most recent year available. In all specifications coefficients are negative and statistically significant, thus providing further support for the findings in Goel et al. (2012) that corruption perceptions decrease as Internet awareness increases. IV. Conclusion Overall, the results suggest that the effect Internet awareness has on corruption is highly sensitive to the corruption measures used and the time the Internet data is constructed. It seems

6 that corruption perceptions are highly influenced by Internet awareness, while the effect Internet awareness has on corruption experience in unclear. This paper finds that Internet awareness decreases corruption perceptions and corruption experience of households, while it increases the frequency of corruption experience in firms.

7 References Bologna, J Corruption: the good, the bad, and the uncertain. Working paper. Brunetti, A., Weder, B A free press is bad news for corruption. Journal of Public Economics 87(7-8), Chowdhury, S.K The effect of democracy and press freedom on corruption: an empirical test. Economic Letters 85, Djankov, S., McLiesh, C., Nenova, T., Shleifer, A Who owns the media? NBER Working Paper No Donchev, D. and Ujhelyi, G What do corruption indices measure? University of Houston Goel, R. K., Nelson, M.A., Naretta, M.A The Internet as an indicator of corruption awareness. European Journal of Political Economy, 28, Kenny, C Measuring corruption in infrastructure: evidence from transition and developing countries. Journal of Development Studies, 45, Mocan, N What determines corruption? International evidence from microdata. Economic Inquiry, 46, Treisman, D What have we learned about the causes of corruption from ten years of crossnational empirical research? Annual Review of Political Science 10, Weber Abramo, C How much do perceptions of corruption really tell us? Economics: The Open-Access, Open-Assessment E-Journal 2 (2008-3).

8 Table 1: Variable descriptions and summary statistics. Variable Description (year; mean; std. dev.) Source INTERNET_HITS Average number of Internet search hits per-capita (2013; 2.468; 8.069). CPI Corruption perceptions index; range 0-10, higher value, more corrupt (2008; 5.781; 2.140). TI CC Control of corruption index; range -2.5 to 2.5; higher value, more corrupt (2007; ; 1.004). WB APAY Index of percent of firms that paid a bribe; range 1-6; higher value, more corrupt. (2000; 2.879; 0.854). WB BRIBE_WBES Percent of total sales paid in bribes (2000; 2.987; 2.137). WB BRIBE_UNICRI Percent of household respondents that paid a bribe (2000; 9.408; 8.855). UNICRI; Mocan, 2008 BRIBE_GCB Percent of respondents that paid a bribe (2005; ; ). TI GDP_PC GDP per-capita; PPP adjusted; 2005 constant dollars (2007; 13, ; 15, ) WB G_GDP Government consumption as percent of GDP (2006; ; 5.482). WB FREEDOM Index of economic freedom; range 0-100; higher value, more freedom (2009; ; ). Heritage Foundation VOICE Voice and accountability index; range -2.5 to 2.5; higher value, stronger government (2008; ; 0.992) WB EFFECTIVE Government effectiveness; range -2.5 to 2.5; higher value, stronger government (2008; 0.074; 0.957). WB RULE Rule of law; range ; higher value, stronger government (2008; ; 0.981). WB Notes: Summary statistics are given for the year included in the Goel et al. (2012) paper.

9 Table 2: Internet corruption awareness and its effect on corruption. Panel A: (dependent variable: CPI_2008) 2A.1 2A.2 2A.3 2A.4 INTERNET_HITS *** *** *** *** (0.004) (0.004) (0.006) (0.004) Observations R Panel B: (dependent variable: CC_2007) 2B.1 2B.2 2B.3 2B.4 INTERNET_HITS *** ** *** *** (0.003) (0.003) (0.003) (0.002) Observations R Panel C: (dependent variable: APAY) 2C.1 2C.2 2C.3 2C.4 INTERNET_HITS 0.025*** 0.024*** 0.024*** 0.023*** (0.003) (0.003) (0.003) (0.003) Observations R Notes: *, **, *** denote statistical significance at the 1, 5, and 10 % levels, respectively. Robust standard errors in parenthesis. Constant term included in all regressions. Full set of controls included.

10 Table 3: Internet corruption awareness and its effect on alternative measures of corruption. Panel A: (dependent variables: BRIBE_WBES) 3A.1 3A.2 3A.3 3A.4 INTERNET_HITS Observations R Panel B: (dependent variable: BRIBE_UNICRI) 3B.1 3B.2 3B.3 3B.4 INTERNET_HITS ** * Observations R Panel C: (dependent variable: BRIBE_GCB) 3C.1 3C.2 3C.3 3C.4 INTERNET_HITS Observations R Notes: *, **, *** denote statistical significance at the 1, 5, and 10 % levels, respectively. Robust standard errors in parenthesis. Constant term included in all regressions. Full set of controls included.

11 Table 4: Internet corruption awareness and its effect on corruption perceptions using updated data. Panel A: (dependent variable: CPI_2012) 4A.1 4A.2 4A.3 4A.4 INTERNET_HITS *** ** *** *** (0.003) (0.003) (0.002) (0.002) Observations R Panel B: (dependent variable: CC_2012) 4B.1 4B.2 4B.3 4B.4 INTERNET_HITS ** * *** ** (0.002) (0.002) (0.001) (0.001) Observations R Notes: *, **, *** denote statistical significance at the 1, 5, and 10 % levels, respectively. Robust standard errors in parenthesis. Constant term and full set of controls included in all regressions.