Mobile Phone Diffusion and Corruption in Africa

This article was downloaded by: [Bailard, Catie Snow] On: 4 August 2009 Access details: Access Details: [subscription number 913658677] Publisher Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Political Communication Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713774515 Mobile Phone Diffusion and Corruption in Africa Catie Snow Bailard a a School of Media and Public Affairs, George Washington University, Online Publication Date: 01 July 2009 To cite this Article Bailard, Catie Snow(2009)'Mobile Phone Diffusion and Corruption in Africa',Political Communication,26:3,333 — 353 To link to this Article: DOI: 10.1080/10584600903053684 URL: http://dx.doi.org/10.1080/10584600903053684 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. Political Communication, 26:333–353, 2009 Copyright © Taylor & Francis Group, LLC ISSN: 1058-4609 print / 1091-7675 online DOI: 10.1080/10584600903053684 Mobile Phone Diffusion and Corruption in Africa 1091-7675 1058-4609 UPCP Political Communication, Communication Vol. 26, No. 3, Jun 2009: pp. 0–0 Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Corruption Catie Snowin Bailard Africa CATIE SNOW BAILARD The explosion of mobile phones into a region that, until recently, was nearly devoid of telecommunications infrastructure provides a valuable opportunity to explore the potential effects of information and communication technology on various economic and social outcomes. This article focuses specifically on the potential influence that mobile phones will exert on corruption in Africa. Two distinct empirical analyses test the hypothesis that mobile phones will reduce corruption in Africa, as a result of decentralizing information and communication and thereby diminishing the opportunities available to engage in corruption as well as increasing the potential of detection and punishment. The results of a fixed effects regression of panel data at the country level reveal a significant negative correlation between a country’s degree of mobile phone penetration and that country’s level of perceived corruption. In addition to this, a multivariate regression of survey data reveals that the degree of mobile phone signal coverage across 13 Namibian provinces is significantly associated with reduced perceptions of corruption at the individual level. Keywords corruption, Africa, mobile phones, ICT, information and communication technologies A woman provides for her family by selling the fish she catches from the river. She does not own a freezer. So, if she can not find customers quickly enough, the fish putrefy and an entire day of work is wasted. Then she buys a mobile phone. Customers can now call her to tell her if they plan to buy her fish that day, which she keeps tethered in the river until they arrive. Now she knows exactly how many fish she is likely to sell that day, enabling her to provision her time and energy more efficiently (LaFraniere, 2005). A 45-year-old farmer needs to know how much to sell her fruit for on a given day. Some days she spends hours trudging around in pursuit of a working public pay phone to call the markets and learn how to best price her fruit. Then the farmer buys a mobile phone. Now, not only can she quickly learn the best price to maximize that day’s sales, but she can also reliably and instantly connect with customers, brokers, and the market (Ngowi, 2005). A man wants to open a bank account. Keeping money in his home makes his family a target for robbery, and it is simply too difficult to amass any sort of savings that way. But Catie Snow Bailard is Professor in the School of Media and Public Affairs at George Washington University. The work for this article was completed while I was a PhD candidate in the Political Science Department at the University of California, Los Angeles. For feedback and guidance with the article and/or beyond, I would like to thank Matt Baum, Tim Groeling, Jeff Lewis, James Lo, Dan Posner, Phil Potter, Lynn Vavreck, and John Zaller. Of course, any shortcomings belonging to the article are entirely my own. Address correspondence to Catie Snow Bailard, School of Media and Public Affairs, George Washington University, 805 21st St. NW, Suite 400, Washington, DC 20052. E-mail: [email protected] 333 Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 334 Catie Snow Bailard there is no bank branch in his remote village. And, even if there were, he surely could not afford the fees associated with a traditional account. Then the man gets a mobile phone. Using his mobile phone and a government-issued identity number, the man calls into a mobile phone banking system. Now it doesn’t matter that there is no bank branch for many miles around. And, with no start-up fees or monthly charges (only a small fee per transaction), the man can now afford to keep his money in a bank account. Not only is his family’s safety and savings less vulnerable to would-be predators, but he can also buy groceries and conduct a number of other daily business transactions with a simple touch of a button (Itano, 2005). Mobile phones are changing daily life in Africa. Anecdotes are already pouring out of the continent, highlighting the many ways that mobile phones have reconfigured and streamlined various daily activities—particularly those in the economic domain. As these real-life examples attest, thanks to mobile phones, buyers and sellers can now connect more readily, creating more efficient markets. Small business owners and individual vendors are also now less dependent on middlemen, reducing their susceptibility to extortion or simply bad information (Ross 2004). And bank accounts via mobile phones render business transactions less vulnerable to default by reducing the risk of nonpayment. As such, these anecdotes are substantiated by a number of recent studies, which have identified mobile phones as contributing to reduced price discrepancies (Aker, 2008), increased consumer and producer welfare (Abraham, 2007; Jensen, 2007), improved productivity (Donner, 2005; Lane, Sweet, Lewin, Sephton, & Petini, 2006; Moloney, 2005), and assistance in finding employment (Frost & Sullivan, 2006; Samuel, Shah, & Hadingham, 2005). Considering the abject poverty still suffered across the African continent, the full range of avenues through which mobile phones may reform and energize African economies is a worthwhile consideration. Accordingly, this article focuses on the mobile phone’s impact within a specific domain of economic activity: corruption. In the following sections, I review the growth of mobile phones and current state of corruption in Africa, followed by a discussion of the impact that I hypothesize mobile phones will exert on corruption in this region. In short, I argue that the net effect of the diffusion of mobile phones in Africa will be to reduce corruption by decentralizing information and communication, thereby diminishing the opportunities available to engage in corruption as well as increasing the potential of detection and punishment. I then test this hypothesis by means of two distinct empirical analyses. First, I employ a fixed effects regression of panel data to test whether the growth in mobile phone penetration across 46 African countries, from 1999 to 2006, is correlated with changes in these countries’ perceived corruption scores.1 Next, in order to better address the endogeneity and misspecification concerns that accompany such cross-country quantitative analyses, I shift the level of analysis to the individual. To do so, I utilize Afrobarometer survey data to determine whether moving along the spectrum from more to less mobile phone signal coverage exerts a distinct and significantly more negative influence on corruption perceptions across the 13 different Namibian provinces in 2006 (when mobile phones were present) compared to 1999 (before mobile phones were widely in use). The findings produced by each of these analyses support the hypothesized negative relationship between mobile phone diffusion and corruption in Africa. The fixed effects regression of panel data reveals a significant negative correlation between a country’s degree of mobile phone penetration and its level of perceived corruption. In addition, the individual-level regression reveals a significantly greater negative correlation between the degree of mobile signal coverage in a respondent’s province and that respondent’s perception of corruption in 2006 when compared to 1999. Corruption in Africa 335 Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Mobile Telephones in Africa As recently as 2000, less than 3% of Africans subscribed to fixed telephone land lines. As a result, mobile telephone companies initially overlooked the African continent. Assuming that, whether it be a result of financial limitations or simple lack of interest, the apparent paucity of demand for land lines among Africans would extend to mobile telephones as well. “We all misread the market,” admits the chief executive officer of a Kenyan mobile service provider (Ngowi, 2005). Despite the number of land-line telephone subscribers continuing to stagnate at less than 4%, the number of mobile phone subscribers mushroomed from less than 2% in 2000 to nearly 30% in 2007 (International Telecommunications Union [ITU], 2008). And, being that this indicator measures only subscribers—defined as prepaid and postpaid users of “portable phones subscribing to an automatic public mobile telephone service” (ITU, 2008, p. 125)—the number of individuals with access to mobile phones likely reaches considerably higher once neighbors, friends, and family are taken into account. Considering that some remote villages still communicate by beating drums, while in other cases merely passing on a message often meant that “someone may have to leave work, travel for days, spending much more money” (LaFraniere, 2005), this rapid infusion of mobile phones into Africa suggests a revolution in communication and information diffusion. As it turns out, the lack of land-line subscriptions was not at all a result of lack of interest in telephonic communication. “The real reason for weak demand was that land lines were expensive, subscribers had to wait for months get hooked up, and the lines often went down because of poor maintenance, floods, and theft of copper cables” (Ngowi, 2005). As such, once cash-strapped governments began to privatize their telecommunications industries in the mid-1990s, monopolies gave way to more competitive industries that offered increasingly affordable and reliable mobile services. Considering the myriad ways that mobile telephones have already demonstrably improved daily life on the continent—from conducting simple business transactions, to calling a doctor when a baby gets sick in the middle of the night, to being able to check in on loved ones living far away or in war zones—experts do not expect demand to abate any time soon (Sullivan, 2006). In fact, at the moment, it is the supply side that is struggling to keep pace. Mobile towers cannot be erected quickly enough. In Nigeria, two providers had to suspend sales of subscriber identity module (SIM) cards for a period of months to allow the companies to strengthen their networks. In the Congo, unwilling to wait for the mobile company to erect a nearby mobile tower, eager villagers built 50-foot-high treehouses to capture signals from remote towers (LaFraniere, 2005). However, villagers such as these may not have to wait much longer: As of 2007, mobile phone signal coverage reached 64% of the African population (ITU, 2008). Moreover, as mobile subscriber rates near 100% in developed countries, although the majority of future African subscribers are of low income, providers are nevertheless keen to get them onboard. “Existing business models and strategies show that lower revenues are compensated by masses of new subscribers. Also, studies have shown low-income groups are prepared to spend proportionally more of their income on telecommunications” (ITU, 2008, p. 6). As such, mobile providers are producing increasingly affordable mobile handsets and selling airtime in cheaper and cheaper packets, with the majority of subscribers opting for prepaid plans that offer such features as per-second billing and low denomination airtime recharges (ITU, 2008). 336 Catie Snow Bailard Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Corruption in Africa Corruption is widely defined in the literature as behavior that “deviates from the formal duties of a public role because of private-regarding (personal, close family, private clique) pecuniary or status gains; or violates rules against the exercise of certain private-regarding behavior” (Klitgaard, 1988, p. 23). Corruption is not unique to Africa, but the extent to which it has hamstrung good governance and economic development may be. As one scholar observes, corruption “is not markedly worse than in many other parts of the developing and former communist world. . . . Yet corruption in Africa is universally perceived, by external observers and by local reformers alike, as being ‘catastrophic’ in its impact on development” (Szetfel, 2000, p. 427). In brief, scholars have identified corruption as contributing to fiscal deficits, income inequality, market distortion, lack of investment, inefficient public services, low growth rates, authoritarianism, political instability, and state collapse (Tanzi, 1998; Mauro, 1995, 1997; Szetfel, 2000; Azfar, Lee, & Swamy, 2001; Jackson & Rosberg, 1994). In Africa, corrupt behavior often takes the form of rent-seeking, in which elites and officials extract payments in return for special policy, administrative advantages, or other political goods (LeVine, 1975). Aside from these payments, which Western society generally deems “bribes,” corruption also takes the more passive form of officials and local elites siphoning away some portion of aid and welfare disbursements for their own private use. For example, a study tracking the distribution of school capitation grants in Uganda found that only 14% of the funds intended for the schools actually reached their targets (Reinikka & Svensson, 2004). Similar studies reveal comparable levels of aid lost through misappropriation in Ghana (49%), Tanzania (57%), and Zambia (76%) (Ye & Canagarajah, 2002; Price Waterhouse Coopers, 1999; Das, Habyarimana, & Krishnan, 2002; Reinikka & Svensson, 2004). In African nations, the funds derived from corruption frequently fuel clientelist politics (Jackson & Rosberg, 1994; Reinikka & Svensson, 2004), which are defined as “the distribution of public goods—offices, public works projects, permits, tax breaks, and so on—in return for loyalty” (Leonard & Scott, 2003, p. 2) On the ground level, clientelist politics often also entail vote-buying, in which small monetary payments or staples are provided to villagers in exchange for their vote commitment. However, portions of these misappropriated funds also ultimately line the pockets of officials and elites (Platteau & Gasspart, 2003). While there tends to be scholarly consensus regarding the deleterious nature of corruption,2 the primary causes of corruption in Africa remain much more debated. Some scholars contend that corruption is an artifact of colonialism, in which African leaders of newly independent states sought to replicate the material indulgences of their former colonial leaders (Greenstone, 1966). However, another line of reasoning contends that corruption derives primarily from precolonial notions of patrimonialism, in which leaders must capture funds to satisfy the expectations of traditional kinship networks and ethnic obligations (Olivier de Sardan, 1999). Turning to more economic-based arguments, some argue that integration into the global market is correlated with decreased corruption as a result of exposing domestic businesses to international norms and increased demands for transparency (Sandholtz & Koetzel, 2000; Mauro, 1995). However, another school of thought contends that increased participation in the global economy actually encourages corruption by creating more situations in which companies seeking contracts will find the payment of bribes to be highly lucrative (Tanzi, 1998). Moving to politics, one body of literature contends that more competitive political systems and democracy in general diminish corruption by increasing accountability and transparency (Rose-Ackerman, 1993; Rasmusen & Ramseyer, 1994; Montinola & Jackman, 2002). Conversely, Szetfel (2000) worries that Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Corruption in Africa 337 intense factional competition for political office could escalate political mobilization via patronage politics, increasing the incentives for leaders to engage in corrupt activities. Despite these very divergent arguments, which represent only the tip of the iceberg in terms of the range of factors argued to contribute to corruption, there are two simple logical premises that tend to underpin most theories regarding the causes of corruption. First, corrupt behavior most often takes place in secrecy. Corruption prefers the shadows and abhors transparency, so corruption will thrive in contexts where information and communication are centralized into a few, often elite, hands (Reinikka & Svensson, 2004, 2005; Schleifer & Vishny, 1993; Sandholtz & Koetzle, 2000). This preference for privacy stems from the second logical premise, which posits that individuals will engage in a costbenefit analysis in the course of deciding whether to commit an act of corruption. In essence, individuals weigh the likely benefit to be gained from the corrupt act against the potential cost of committing the act—particularly, the likelihood of detection and punishment (Azfar et al., 2001; Schleifer & Vishny, 1993; Bardhan & Mookherjee, 2000; Treisman, 2000). Taking these two premises into account, it is not surprising that corruption has proliferated on the African continent, where telecommunication technologies— the very tools that promote transparency and strengthen oversight mechanisms—were, until recently, nearly nonexistent among the general population. Theory Although a number of institutional reforms are necessary to effectively curb corruption, transparency and accountability are consistently cited as integral components of anticorruption campaigns. Transparency and accountability combat corruption both by decreasing the number of opportunities available to engage in corruption and increasing the likelihood of detection and punishment. The historic dearth of telecommunications infrastructure in Africa centralized information and communication into the hands of elites, creating conditions conducive to the perpetration of corrupt behavior by severely constraining the potential for transparency and accountability. As such, I argue that the net effect of the rapid and massive diffusion of mobile phones in Africa will be the reduction of corruption by decentralizing information and communication, thereby shrinking the veil of secrecy that shields corrupt behavior as well as altering the cost-benefit calculus of corrupt behavior by strengthening oversight and punishment mechanisms. In order to engage in corrupt behavior, the opportunity to do so must first exist. Therefore, it can be plausibly reasoned that the more opportunities that are available, the more likely individuals will engage in corrupt behavior. Opportunities for corruption increase when elites, officials, and other influential individuals enjoy some degree of monopoly over information and communication. For example, local elites can capture aid disbursements intended for schools or villages because they represent the exclusive channel for communication between aid agencies and the intended recipients—enabling local elites to falsely report back to aid agencies as well as keep recipients in the dark regarding the actual amount of intended aid (or the existence of such aid, at all) (Reinikka & Svensson, 2004, 2005). Bureaucrats and officials can extract bribes because they enjoy sole discretion over the supply of required services, permits, or licenses (Azfar et al., 2001). Moreover, even in instances when officials do not officially possess monopoly discretion, they can exploit villagers’ comparative lack of information regarding the appropriate costs and full range of channels available to procure such services, permits, or licenses. Mobile phones decrease the prevalence of opportunities for corruption by decentralizing information and communication. Thanks to mobile phones, local elites no longer Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 338 Catie Snow Bailard necessarily serve as the exclusive channel for communication between aid agencies and intended recipients. With mobile phones, aid agencies can directly contact schools and villagers to ensure that aid was appropriately disbursed. In addition, mobile phones make it easier for villagers to learn that they are entitled to receive a certain amount of aid, increasing their capacity to demand that aid. Mobile phones also diminish the power of local officials to extract bribes by better connecting individuals with alternative officials or with villagers who can provide information regarding alternative channels, reducing a given official’s sole discretion over the supply of services, permits, or licenses. In support of the power of information diffusion to deter local capture, a recent study revealed that simply posting a newspaper advertisement stating that aid was meant to be dispersed to certain schools significantly reduced the amount of aid that was lost through misappropriation (Reinikka & Svensson, 2005). Notably, this study also found that the farther a school was geographically located from the closest newspaper outlet, the more funding that school lost to local capture. Not only do mobile phones serve as an additional channel for informing villagers regarding intended aid, adding mobile phones will further strengthen newspaper public information campaigns against corruption by reducing the constraining effect of geography and literacy. In addition to these instances, mobile phones can also directly reduce the opportunities for corruption as a result of the spread of affordable bank accounts via mobile phones. As more Africans open bank accounts via mobile phones, aid agencies and governments are beginning to transfer aid, pensions, and grants directly into the bank accounts of recipients (Itano, 2005). As a result, some degree of aid disbursement will bypass local elites entirely, meaning that the opportunity to capture these funds should diminish considerably. Moving to the next stage, mobile phones also reduce corruption by increasing the likelihood of detection and punishment, thereby changing the cost-benefit calculus of corrupt behavior. In the past, the dearth of telecommunications infrastructure meant that aid agencies and officials had to rely on local elites for assurances that aid was disbursed appropriately. If suspicious of the veracity of these reports, their only alternative often was embarking on the arduous and timely endeavor of traveling to remote villages to verify in person. Accordingly, it was often difficult to definitively determine whether corruption had occurred at all, let alone punish it. In addition, although many African governments have instituted (often sincere) anti-corruption reforms, “experience has shown that these need to be backed up with effective enforcement abilities, especially the ability to track down violations” (Harsch, 1993, p. 46)—a rather difficult endeavor when control over communication and information is tightly centralized within the same small group of individuals that largely perpetrate the corruption. Mobile phones reduce the monopoly over information and communication that officials have traditionally exploited to shield their own corrupt behavior. In addition to enabling aid agencies to efficiently and directly follow up with intended aid recipients, mobile phones better equip reformers to track down violations and monitor adherence to anti-corruption campaigns. In support of this, a mobile service provider in Sierra Leone launched a toll-free phone number that residents can call to report instances of suspected corruption to the nation’s Anti-Corruption Commission (Ogendeji, 2008). Finally, mobile phones also bolster the role that average citizens can play in the fight against corruption by increasing the ease with which whistle-blowing citizens can contact reformers, government officials, and the news media. The recent conviction of the son of an Indian politician provides an example of how mobile phones can empower citizens in the fight against corruption; in that instance, a massive mobile phone grassroots campaign forced prosecutors to retry a case in which a politician’s son had been previously acquitted despite shooting a bartender dead in front of several eyewitnesses (Rautray, 2006). Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Corruption in Africa 339 Note, however, that I do not claim that mobile phones alone will act as a panacea for corruption. Successful anti-corruption campaigns require a multidimensional approach in which mobile phones can only play a partial role. Therefore, although the diffusion of mobile phones into a region where telecommunications technology was nearly nonexistent may serve to reduce corruption by decentralizing information and communication, there is no reason to expect mobile phones alone to eradicate corruption. After all, there are a number of countries outside of Africa that boast a lot of phones and a lot of corruption. Accordingly, in order for mobile phones to decrease corruption, it is necessary that there be someone on the other end of the line committed to the fight against corruption. If there are no concerned citizens, aid agency representatives, reformers, or journalists “dialing in” in the fight against corruption, phones alone will likely make little difference. Therefore, the presence of some combination of public awareness campaigns, responsible and vigilant aid agencies, sincere reform efforts, and/or a vigilant and unfettered news media seems necessary for mobile phones to exert a negative influence on corruption. Lastly, there exist a number of instances in which the diffusion of mobile phones may actually enhance rather than inhibit corrupt behavior. These instances can be characterized as those in which some segment of the masses are directly involved in and benefit from the corrupt act. Two examples of this category of corrupt acts are vote-buying and ethnic kinship demands. Villagers directly benefit from the foodstuffs and small monetary payments they receive in exchange for pledging their vote to a given candidate. Mobile phones feasibly make it easier for candidates and parties to contact these voters for the purposes of making arrangements for the buying of votes. In addition, mobile phones increase the ease with which villagers can contact their officials once elected into office. This will feasibly make it more efficient for villagers to attempt to extract goods by pressuring their officials to honor traditional kinship ties. Nevertheless, despite the conditional nature of the impact of mobile phone diffusion on corruption and despite the likelihood that mobile phones will increase the feasibility of specific types of corrupt acts, this article contends that mobile phone diffusion will inhibit corrupt behavior more often than facilitate it. In other words, the diffusion of mobile phones will improve the ability to avoid, detect, and/or punish corruption to a greater extent than it will enhance the ability to collude, conceal, and/or execute corrupt acts. This is because a large proportion of corrupt acts in Africa can be characterized as collusion between elites to the detriment of the masses. And, long before the diffusion of mobile phones, elites possessed the telephonic resources to orchestrate and commit those acts. As a consequence, the diffusion of mobile phones should only minimally enhance the capacity for these elites to collude for the purposes of corruption. On the other hand, the massive diffusion of mobile phones into a region where telecommunications infrastructure was nearly nonexistent among nonelites greatly decentralizes information and communication. This weakens elites’ monopoly over communication and information, thereby strengthening the potential for transparency and accountability. This will decrease the number of opportunities available to commit corrupt acts as well as augment the potential to detect and punish corrupt behavior. Accordingly, the following sections test the hypothesis that the net effect of mobile phone expansion will be to diminish corruption in African nations. Country-Level Analysis Methods To test whether the diffusion of mobile telephones correlates with decreased perceived corruption levels in African nations, I employ a fixed effects OLS multiple regression Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 340 Catie Snow Bailard based on panel data spanning from 1999 to 2006. The dependent variable used in this analysis is Transparency International’s Corruption Perceptions Index (CPI), which is a composite measure based on surveys from 3 to 14 different sources of country analysts and resident and nonresident businesspeople.3 The original CPI assigns a score of between 1 and 10 (to the 10th unit) to each nation; with lower numeric scores indicating greater levels of corruption. However, for ease of interpretation, this article employs an inverted version of the CPI score, also ranging from 1 to 10—so that higher numeric scores now represent more corrupt nations. Accordingly, a negative correlation coefficient now indicates that a variable is associated with a decrease in corruption levels, while a positive coefficient represents a positive relationship between that variable and corruption. Opting to use the subjective measure of perceived corruption rather than a more objective measure of actual corruption is motivated by two considerations. First, corruption is culturally variable—for example, what some countries might consider bribes, others may accept as a necessary cost of doing business—rendering it a very difficult concept to define consistently and accurately across borders. Second, as the following example from Transparency International (2007) illustrates, reliable and valid data measuring corruption are difficult to obtain: “Comparing the number of prosecutions, for example, does not reflect actual levels of corruption but the quality of prosecutors” (p. 2). As one corruption scholar remarks, “If corruption could be measured, it could probably be eliminated. In fact, conceptually it is not even clear what one would want to measure”; however, “while there are no direct ways of measuring corruption, there are several indirect ways of getting information about its prevalence in a country” (Tanzi, 1998, pp. 576–577). Accordingly, the preferred measure of corruption utilized by corruption scholars is Transparency International’s subjective CPI measure. CPI does not force each nation to conform to the same cookie-cutter formula of corruption, trusting instead that country experts and other individuals well acquainted with each country are best equipped to determine the extent to which official behavior deviates from accepted national norms. The high level of correlation between the 14 sources used to build the 2007 CPI scores substantiates the reliability of this index, which Pearson’s and Kendall’s rank correlations place at .77 and .62, respectively (Transparency International, 2007). I build the primary independent variable of mobile phone penetration from data provided by the International Telecommunications Union, an agency of the United Nations. This variable measures the number of mobile phone subscribers per 100 persons for each nation. Recall that this indicator only represents subscribers, meaning that the number of individuals with access to mobile phones reaches much higher once the friends, family, and neighbors of mobile subscribers are taken into account. In fact, in some villages individuals have begun using their own mobile phones to start small businesses, renting out their mobile phone for a fee (LaFraneire, 2005). Accordingly, the impact of mobile phone penetration in Africa is best operationalized as the logged value of the percentage of mobile subscribers in each nation. First, as previously stated, since the ITU’s measure of mobile subscribers only gauges the number of subscribers, the number of individuals who have access to these phones and therefore benefit from them is certainly higher than this measure would indicate.4 Second, as network analysts contend, the “value of a telephone network increases with the square of the number of users,” since the n + 1st user benefits from being able to communicate with all n prior users, and each of them gains from being able to communicate with her or him (Daly, 2000, p. 288). In other words, as a result of network externalities, the more people who use mobile phones, the larger the number of people those individuals can communicate with using those phones, thereby exponentially increasing the value of mobile Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Corruption in Africa 341 phones. Because of these two considerations, I operationalize the value of mobile phone penetration as the logged value of the number of subscribers in a given country for this analysis. This model also controls for a handful of factors that are correlated with both mobile phone penetration and corruption.5 First, scholars often cite poverty and low wages as exerting causal influences on corruption, since individuals’ need and willingness to capture funds via corrupt behavior increases the greater their level of deprivation (Sandholtz & Koetzle, 2000). As discussed above, scholars also argue that corruption exerts a reciprocal effect on poverty by limiting economic development. Since poverty is also feasibly correlated with mobile phone penetration, by reducing the number of individuals who can afford mobile phone service, this model incorporates a variable measuring GDP per capita. Second, researchers have in addition identified a correlation between democracy and corruption, with corruption levels declining once nations reach the more complete level of democratization (Montinola & Jackman, 2002; Sandholtz & Koetzle, 2000). A strong positive correlation between democratization and the expansion of telecommunications infrastructure also exists (Kedzie, 1997; Best & Wade, 2005). To account for the potentially confounding influence of democracy, this model includes a measure of the strength of democratic practices available in a nation.6 Third, the degree of privatization of a nation’s telecommunications industry is correlated with both mobile phone penetration and corruption. As the International Telecommunications Union (2008) reports, increased competition in the telecommunications industry has been a primary catalyst in the reduction of mobile prices across Africa, contributing to expanded networks, reduced tariffs, and more affordable service provision. Privatization (or the lack thereof) is also correlated with corruption, since “when the state and its administrative apparatus exercise relatively greater control over the economy, public officials make decisions that determine who will enjoy access to economic resources. . . . Thus bribery, extortion, and kickbacks become viable means of influencing distribution of wealth” (Sandholtz & Koetzle, 2000, p. 37). To control for the correlates of privatization, I include a World Bank governance indicator measuring each nation’s regulatory quality.7 This composite variable captures the degree to which government institutions generally protect and promote market-friendly economic competition by means of a number of factors, including business regulations, tax policy, privatization, and protectionism.8 Finally, this analysis controls for fixed effects. The inclusion of year and country variables offers several important advantages in the testing of panel data. First, it allows the model to control for factors that do not change over time, reducing concern that the model fails to account for influential country-specific factors that are either difficult to identify or difficult to measure. Fixed effects also confine the analysis to within-country variation, as opposed to across-country variation, which is a more valid means of determining the effect of mobile phone penetration on corruption within a country by minimizing the acrosscountry characteristics that are plausibly correlated with corruption (Allison, 2005). Results As predicted, greater mobile phone penetration is associated with lower levels of perceived corruption, reaching the .01 significance level (see Table 1). In substantive terms, Cameroon’s increase in mobile penetration from under 1 subscriber in 1999 to 24 subscribers per 100 individuals in 2006 (an increase approximate to just under two standard deviations) predicts a decrease in that country’s perceived level of corruption of .7 points. Considering that the CPI’s entire scale ranges from 1 to 10, this represents an 8-percentage-point 342 Catie Snow Bailard Table 1 Country-level regression of mobile phone penetration on CPI scores Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Slope Log of mobile phone penetration (base 10, mean = .53, SD = .82) GDP per capita (US$, mean = 1,512, SD = 2,084) Strength of democratic practices (percentile ranking, mean = 33, SD = 21) Regulatory quality (percentile ranking, mean = 35, SD = 19) Intercept −.47 SE 0.1 95% confidence interval −.68 to −.27 .00004 0.00006 −.0001 to .0001 Statistical significance .001*** .550 −.02 0.006 −.03 to −.006 .004*** −.004 0.006 −.02 to .008 .480 9.7 0.3 Note. Number of cases: 219; adjusted r2: 92%. The dependent variable is the inverted CPI score, which is a discrete variable ranging from 1 (lowest possible level of corruption) to 10 (highest possible level of corruption), to the 10th unit. It has a mean of 8 and standard deviation of 1.1. Country and year fixed-effects variables are also included in this regression, the coefficients of which are not included in the table for brevity’s sake. These can be obtained from the author upon request. In this table *** indicates significance at the .01 level. improvement in that country’s perceived level of corruption across the entire spectrum of possible scores. Moreover, being that African nations consistently rank among the most corrupt in the world, the countries constituting this analysis actually span a much more narrow range, from 4.6 to 10. Therefore, a .7-point improvement in a country’s CPI score represents an upgrading of 13 percentage points across the range of cases considered in this analysis. Finally, to test the robustness of these results, I conducted several iterations of the original regression by omitting various combinations of independent variables—the results of each largely analogous to those attained in the primary regression (see Table 2). Individual-Level Analysis Methods Although the country-level regression of panel data substantiates the hypothesized negative effect of mobile phone penetration on perceived corruption, an individual-level analysis of this relationship will provide a more compelling empirical foundation for this analysis. As is often the case for cross-national analyses, the country-level analysis alone cannot adequately assuage a number of potential methodological concerns, particularly endogeneity (i.e., when the independent and dependent variables share a reciprocal relationship), omitted variable bias (i.e., failing to include an independent variable that should be in the model), and ecological fallacy (i.e., an error of inference caused by assuming that an observed association between given variables at the aggregate level also exists at the individual level). Therefore, I employ Afrobarometer survey data and mobile phone signal coverage rates across the 13 Namibian provinces to determine whether mobile phone penetration is also correlated with reductions in perceived corruption at the individual level. Corruption in Africa 343 Table 2 Robustness check of country-level regression of mobile phone penetration on CPI scores Log of mobile phone penetration GDP per capita Regression 1 Regression 2 Regression 3 Regression 4 −.52 (SE = .09, p ≤ .01***) −.55 (SE = .1, p ≤ .001***) −.47 (SE = .09, p ≤ .01***) −.53 (SE = .09, p ≤ .01***) .00001 (SE = .00007, p ≤ .88) Strength of democratic practices −.02 (SE = .006, p ≤ .01***) Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Regulatory quality Intercept −.01 (SE = .01, p ≤ .11) 9.0 (N = 277, 8.9 (N = 219, adj. r2 = 91%) adj. r2 = 92%) 9.4 (N = 224, adj. r2 = 92%) 9.3 (adj. r2 = 91%) Note. The dependent variable is the inverted CPI score, which is a discrete variable ranging from 1 (lowest possible level of corruption) to 10 (highest possible level of corruption), to the 10th unit. It has a mean of 8 and standard deviation of 1.1. Each of these regression models includes only the variables for which a correlation coefficient is listed in the table. Country and year fixed-effects variables are also included in each of these regressions, the coefficients of which are not included in the table for brevity’s sake. These can be obtained from the author upon request. In this table *** indicates significance at the .01 level. Of the 12 countries surveyed in both 1999 and 2006, I selected Namibia for this analysis because it was the best fit for three crucial criteria. First, the surveys in 1999 and 2006 both included questions pertaining to corruption with reasonably similar wording. Second, Namibia experienced neither too little nor too great mobile phone signal expansion during this period, permitting adequate variation in the measure of mobile phone signal coverage across the provinces. And, third, the geographical size of the provinces is not so small as to invite concern regarding the pixel quality of the GSM map used to build the independent variable of mobile signal coverage. Factors related to both mobile phone penetration and corruption also render Namibia an attractive candidate for this analysis. First, the growth in mobile phone penetration in Namibia during this period was sufficient to make any causal claims regarding the impact of mobile phones on corruption plausible. In 1999 less than 2% of Namibians subscribed to mobile phones, whereas by 2005 this number had reached nearly 25%. Moreover, the presence of a relatively active Anti-Corruption Commission in Namibia means that this country meets at least one of the conditionality criteria set forth in the theory section.9 I derive the dependent variables measuring corruption perceptions in Namibia from questions asked in the 1999 and 2006 Afrobarometer surveys. The first question asks respondents to indicate how corrupt they believe the president’s administration to be. The second question asks respondents to rate the degree of corruption among local government officials. The final question inquires whether the respondents have had to pay a bribe in order to receive household services, such as water or electricity, over the past year (see Table 3 for exact question wordings). As a whole, these questions paint a relatively inclusive picture of corruption perceptions by tapping into assessments of corruption at both the Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Table 3 Wording of questions from 1999 and 2006 Afrobarometer surveys Perception of corruption in presidential administration 344 Perception of corruption among local government officials Paid a bribe in the past year 1999 2006 What about corruption? (Corruption is where those in government and the civil service take money or gifts from people and use it for themselves, or expect people to pay them extra money or a gift for them to do their job.) How many officials in the Nujoma administration do you think are involved in corruption? (Answers: 1. None, almost none; 2. A few, some; 3. Most; 4. All, almost all; 5. Don’t know) What about corruption? How many officials in your local government do you think are involved in corruption? (Answers: 1. None, almost none; 2. A few, some; 3. Most; 4. All, almost all; 5. Don’t know) How many of the following people do you think are involved in corruption, or haven’t you heard enough about them to say: the president and officials in his office? (Answers: 1. None; 2. Some of them; 3. Most of them; 4. All of them; 5. Don’t know) How many of the following people do you think are involved in corruption, or haven’t you heard enough about them to say: local government officials? (Answers: 1. None; 2. Some of them; 3. Most of them; 4. All of them; 5. Don’t know) In the past year, have you or anyone in your family had to pay money to government officials (besides paying rates or taxes), give them a gift, or do them a favor in order to get the following: electricity or water? (Answers: 1. No; 2. Once or twice; 3. A few times; 4. Often; 5. Don’t know) In the past year, how often, if ever, have you had to pay a bribe, give a gift, or do a favor to government officials in order to get a household service (like piped water, electricity, or phone)?a (Answers: 1. No experience with thisb/no; 2. Once or twice; 3. A few times; 4. Often; 5. Don’t know) a The difference in the wordings of the bribery questions is potentially problematic. Whereas the 1999 version references family members, the 2006 question does not. The inclusion of family members could potentially bias this dependent variable in favor of my hypothesis, since taking into account more people increases the chances of having paid a bribe. On the other hand, the 2006 question asks about bribes paid for water and electricity but also includes telephone service as an example, while the 1999 question asks exclusively about water and electricity. The inclusion of additional services could feasibly bias this variable in the opposite direction than that predicted by my hypothesis, since considering additional services increases the number of opportunities in which that individual may have been bribed. In sum, then, these two opposing potential biases may ultimately counterbalance one another. Moreover, since this regression primarily seeks to determine whether living in a province with greater signal coverage has a distinctly more negative relationship with corruption perceptions in 2006 than it does in 1999, these questions are adequate to permit this analysis. Essentially, the matter of interest is whether moving along the spectrum of less signal coverage to more signal coverage decreases the likelihood of being bribed more dramatically in 2006 (when cell phones were present) than in 1999 (when cell phones were not present). b The 2006 question provided an additional option of answering “No experience with this in past year,” while the 1999 question did not include this option. As such, these answers were collapsed into the “no” category in the 2006 data, since “no” is what respondents in 1999 would plausibly answer if they did not have the option of answering “No experience with this in the previous year.” Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Corruption in Africa 345 national and local levels, as well as the respondents’ actual experience with a particular act of corruption. Since the Afrobarometer does not ask respondents specifically about their access to a mobile phone, the primary independent variable in this model is a measure of the percentage of each Namibian province that was covered by a mobile phone signal in late 2005.10 This is based on the logical assumption that the likelihood an individual uses a mobile phone is correlated with the likelihood that the geographic area in which he or she resides receives a mobile phone signal. Using province-level mobile signal coverage as the independent variable also enables this analysis to address endogeneity concerns that the country-level regression could not resolve. Endogeneity occurs when the independent and dependent variables in a regression share a simultaneous relationship, which would be the case if mobile phones and corruption exert a reciprocal influence on one another. This is problematic for an OLS regression because the regression coefficient becomes biased and inconsistent, making it difficult to measure or validate the independent and direct influence of mobile phone diffusion on perceived corruption. It is preferable, therefore, to identify an instrument that is correlated with the independent variable of interest but uncorrelated with the dependent variable. In this case, this factor should be correlated with mobile phone diffusion but not directly correlated with corruption. Mobile signal coverage arguably meets each of these criteria. First, as discussed above, the logic connecting mobile phone penetration with mobile signal coverage is evident. Second, the findings of a recent econometric analysis (Buys, Dasgupta, Thomas, & Wheeler, 2008) conducted by the World Bank provide a strong basis for arguing that mobile signal coverage at the province level is uncorrelated with corruption. This study reveals that the factors that are most determinant of the presence of cell towers in a specific region include size of adjacent population, costs of maintaining and installing the towers (i.e., accessibility), and the national competition policy. Of these factors, the variable that is most plausibly correlated with corruption is the nation’s competition policy, since market competition and corruption tend to share a negative relationship. However, being that Namibia’s competition policy is determined at the national level and, therefore, applies uniformly to all of its provinces, the competitive climate of the telecommunications industry is likely constant across the provinces included in this analysis. This suggests that the degree of mobile signal coverage at the province level is most driven by population density and accessibility factors, and therefore exogenous to the relationship shared by competition policy and corruption.11 To construct the measure of mobile signal coverage, I overlaid a 50-by-50 grid per 10-degree latitude-longitude square over a 2005 map of mobile phone signal coverage across the Namibian provinces.12 Using this map, I divide the number of grid squares covered by mobile phone signals by the total number of grid squares that each province encompasses. This produces a measure of the percentage of each province’s geographical space in which it was possible to use a mobile phone in 2005. The average level of mobile signal coverage across these 13 provinces is 42% (SD = 25%), with a minimum of 6% in Okavango and a maximum of 85% in Khomas. In order to test whether gaining access to mobile phones altered respondents’ perceptions of corruption in Namibia, I conducted an OLS multiple regression of this 2005 measure of mobile signal coverage on the respondents’ perceptions of corruption in 1999 and 2006. By including an interaction term representing mobile signal coverage in 2006 specifically, it is possible to test whether living in a state with a high degree of mobile signal coverage was correlated with a greater decline in perceived corruption in 2006 Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 346 Catie Snow Bailard (when mobile phones were actually present) than in 1999 (when mobile phones were not yet being widely used). Moreover, this interaction term enables the analysis to address omitted bias concerns that there may be some factors unaccounted for at the province level, which are really driving the proposed relationship shared by mobile phone signal coverage and corruption perceptions. If this is the case, the analysis should reveal a similar correlation between the degree of mobile signal coverage in a province and the level of corruption perceived by its residents in both 1999 and 2006. On the other hand, if mobile phones do exert an independent influence on corruption perceptions, the correlation between mobile signal coverage and corruption perceptions should be distinct from and significantly more negative in 2006 than in 1999. Finally, this analysis includes several demographic variables to control for personal characteristics that are plausibly correlated with both the likelihood of using a cell phone and attitudes regarding corruption. These factors include gender, age, education, employment status, and whether the individual lives in an urban or rural area. Lastly, the model includes dummy variables for each of the 13 Namibian provinces to control for difficult to identify or difficult to measure province-level factors. Results In all three cases, living in a province with greater mobile signal coverage exhibited a significantly more negative relationship with corruption perceptions in 2006 when compared to 1999. In other words, moving along the spectrum of more to less mobile signal coverage is associated with a more significant decline in corruption perceptions in 2006 (when mobile phones were actually present) than in 1999 (before mobile phones were widely present). Moreover, in each case, the distinct 2006 relationship shared by mobile signal coverage and corruption perceptions reaches statistical significance at the .01 level (see Table 4). In substantive terms, in 1999 residents of the province of Otjozundjupa (mobile signal coverage: 60%) rated the presidential administration to be 3 percentage points less corrupt than did residents of Hardap (mobile coverage: 40%). By 2006, residents of Otjozundjupa rated the extent of corruption in the presidential administration to be nearly 10 percentage points lower than did Hardap residents. The difference in the extent of corruption that respondents assigned to their local governments followed a similar trajectory between 1999 and 2006. In 1999, the level of corruption that residents assigned to their own local government was 4 percentage points lower in Otjozundjupa (mobile signal coverage: 60%) than in Karas (mobile signal coverage: 31%). In 2006, this difference expanded three-fold, to 14 percentage points. Turning to the matter of bribery, individuals living in the province of Oshikoto (mobile signal coverage: 83%) were 5 percentage points less likely to pay a bribe for household services in 1999 than respondents living in the province of Kavango (mobile signal coverage: 6%). By 2006, however, Oshikoto residents had become 15 percentage points less likely to pay a bribe for household services than Kavango residents. In summary, in all three cases, the effect of moving along the spectrum from lower to higher mobile signal coverage on corruption perceptions was significantly more negative in 2006 than in 1999 (see Figures 1, 2, and 3). When compared to 1999, residents living in provinces with a greater degree of mobile signal coverage were significantly less likely to have paid a bribe for household services than individuals living in provinces with less mobile coverage in 2006. The effect of living in a province with higher signal coverage also diminished the level of corruption that respondents assigned to the presidential administration and local government to a greater extent in 2006 when compared to 1999. Taken collectively, these results provide a Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Table 4 Individual-level regression of mobile phone signal coverage on perceptions of corruption Corruption in presidential administrationa 347 Mobile signal coverage of province (0–1 range, mean = .42, SD = .25) Year 2006 (0–1 binary, mean .5, SD = .5) Interaction of mobile signal coverage and year 2006 (0–1 range, .22, SD = .29) Male (0–1 range, mean = .49, SD = .5) Age (0–1 range, mean =.25, SD = .29) Education (0–1 range, mean = .36, SD = .29) Employment status (0–1 range, mean = .34, SD = .44) Urban (0–1 range, mean = .37, SD = .48) Intercept Corruption among local government officialsb Had to pay bribe in past year for household servicesc −.17 (SE = .15, p ≤ .26 ) −.13 (SE = .82, p ≤ .87) −.07 (SE = .48, p ≤ .89) .02 (SE = .03, p ≤ .41) −.31 (SE = .05, p ≤ .001***) .22 (SE = .03, p ≤ .001***) −.34 (SE = .06, p ≤ .001***) .11 (SE = .01, p ≤ .001***) −.10 (SE = .03, p ≤ .005***) .01 (SE = .01, p ≤ .29) −.03 (SE = .03, p ≤ .23) −.02 (SE = .03, p ≤ .49) −.03 (SE = .02, p ≤ .07*) .020 (SE = .01, p ≤ .27) .002 (SE = .03, p ≤ .95) .006 (SE = .03, p ≤ .83) −.030 (SE = .02, p ≤ .06*) .005 (SE = .007, p ≤ .47) −.030 (SE = .01, p ≤ .03**) −.040 (SE = .01, p ≤ .004***) .020 (SE = .009, p ≤ .06*) .04 (SE = .02, p ≤ .03**) .57 (N = 1,810, adj. r2 = 13%) .04 (SE = .01, p ≤ .08*) .47 (N = 1,699, adj. r2 = 8%) .008 (SE = .01, p ≤ .43) .06 (N = 2,128, adj. r2 = 5%) Note. For the sake of brevity, I am not reporting the coefficients of the province dummy variables, but these results can be obtained from the author. In this table * indicates significance at the .1 level, ** indicates significance at the .05 level, and *** indicates significance at the .01 level. a The dependent variable is how many members of the presidential administration respondents believe are corrupt. It ranges from 0 (none) to 1 (all), with intermediate values at .33 (some) and .66 (most). It has a mean of .35 and standard deviation of .3. b The dependent variable is how many members of the local government that respondents believe are corrupt. It ranges from 0 (none) to 1 (all), with intermediate values at .33 (some) and .66 (most). It has a mean of .38 and standard deviation of .3. c The dependent variable is whether the respondent was asked to pay a bribe for household services in the previous year. It ranges from 0 (no) to 1 (often), with intermediate values at .33 (once or twice) and .66 (a few times). It has a mean of .05 and standard deviation of .17. 348 Catie Snow Bailard Change in Average Perception of Presidential Corruption 20% 0% –20% –40% 1999 2006 –60% 0% 20% 40% 60% 80% 100% Percentage of Province Covered by Mobile Phone Signal Figure 1. Different relative effects of living in a province with greater mobile signal coverage on perceptions of presidential corruption before and after diffusion of mobile phones. 20% Change in Average Perception of Local Government Corruption Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 –80% 0% –20% 1999 2006 –40% –60% –80% 0% 20% 40% 60% 80% 100% Percentage of Province Covered by Mobile Phone Signal Figure 2. Different relative effects of living in a province with greater mobile signal coverage on perceptions of local government corruption before and after diffusion of mobile phones. relatively broad empirical foundation supporting the hypothesized negative influence of mobile phone diffusion on corruption perceptions, by encompassing corruption assessments at both the national and local levels as well as experience with a particular act of corruption. Discussion and Conclusion This analysis substantiates the hypothesized negative influence of mobile telephone diffusion on corruption perceptions at both the country and individual levels. These findings Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Change in Likelihood of Paying Bribe for Water or Electricity Corruption in Africa 349 10% 0% –10% 1999 2006 –20% –30% –40% 0 0.2 0.4 0.6 0.8 1 Percentage of Province Covered by Mobile Phone Signal Figure 3. Different relative effects of living in a province with greater mobile signal coverage on likelihood of paying bribe for water or electricity before and after diffusion of mobile phones. independently and mutually support the prediction that mobile telephone diffusion will decrease corruption in Africa by decentralizing control over information and communication—thereby decreasing the opportunities available to commit corrupt acts as well as increasing the threat of detection and punishment. Nevertheless, there remain some potential caveats that should be considered. To begin with, recall that the dependent variables in this analysis largely represent subjective measures of perceived corruption, as opposed to more objective measures of actual corruption. Opting to focus on perceived corruption was motivated by two considerations. First, corruption is a culturally variable concept, making it very difficult to define consistently and accurately across borders. Second, reliable and valid data measuring “actual” corruption are difficult to obtain. Although scholars generally accept perceived corruption as a valid and reliable proxy measure of actual corruption, it is plausible that mobile phones reduced perceptions of corruption without actually diminishing the incidence of actual corruption. For example, perhaps people conflate cell phones and the ability to communicate with greater transparency, thereby assuming that more cell phones equal less corruption. This suggests that over time, once the gleam wears off, the negative correlation between mobile phone diffusion and corruption perceptions will cease. Or, it is possible that individuals are so pleased that they can now afford to use mobile phones that a sort of halo effect is produced that causes mobile phone users to view their government officials more positively in general. Although these are important potential caveats, I contend that controlling for privatization and GDP per capita at the country level and employment and education status at the individual level should largely control for these possible influences. For example, people’s intellectual capacity to associate mobile phones with greater transparency and, therefore, reduced corruption should be somewhat dependent on their level of education. Second, if individuals are rewarding officials because they are pleased that they can now afford mobile phone service, this potentially confounding influence should be somewhat mitigated by the inclusion of the GDP per capita and privatization measures at the aggregate level, as well as the employment status variable at the individual level. Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 350 Catie Snow Bailard Finally, future analyses should explore the conditionalities upon which I argue a negative relationship between mobile phones and corruption rests. For example, to what degree is some combination of public information campaigns, a vigilant press, sincere reform efforts, and/or responsible aid agencies a necessary requisite for mobile phones to exert a negative net effect on corruption? Moreover, what is the relative influence of each of these factors? In addition, future analyses should also investigate my claim that the Internet will actually increase the incidence of specific types of corrupt acts. I define these particular types of acts as those the masses are directly involved in and gain some immediate benefit from—such as vote-buying and kinship demands. Beyond this, also consider whether the expansion of the telecommunications industry itself will create more opportunities for corruption as the executives of these companies negotiate licenses and contracts with government officials. (However, I would argue that the increase in corrupt behavior within the telecommunications industry specifically would likely be offset by the increased ease with which reformers and journalists could monitor and detect corruption in already-extant industries.) Accordingly, these considerations raise the question of whether mobile phones will instead have the net effect of actually increasing corruption in nations without at least one of the conditions listed above. In conclusion, the findings produced in both the country- and individual-level analyses support the hypothesized negative influence of mobile phone diffusion on corruption in Africa—reaching the .01 significance level in each case. Drawing from 46 nations over the period 1999 to 2006, a fixed effects regression of panel data reveals a significant negative correlation between a country’s degree of mobile phone penetration and its level of perceived corruption. Based on Afrobarometer surveys conducted in Namibia in 1999 and 2006, individual-level regressions also reveal a significantly distinct negative correlation between mobile signal coverage and respondents’ perceptions of corruption in 2006. In more detail, when compared to 1999 (before mobile phones were widely used), living in a province with greater mobile signal coverage in 2006 (when mobile phones were widely in use) significantly reduced how corrupt respondents tended to rate their presidential administration and local government relative to respondents living in provinces with lower mobile signal penetration. Moving along the spectrum from lesser to greater mobile signal coverage also created a more significant reduction in the likelihood of having to pay a bribe for household services in 2006 than it did in 1999. Notes 1. CPI scores and mobile phone penetration rates were not available for each of these countries in each of these years. Therefore, this analysis includes all of the cases in which both measures were available for a given country in a given year—equaling 219 cases in all. For a specific list of these cases, please contact the author. 2. For exceptions to this view, see Huntington (1968) and Leff (1964). 3. For more details about the how this measure is constructed, please visit http://www.transparency. org/ policy_research /surveys_ indices/cpi. 4. Although in some countries subscribers may own multiple SIM cards, potentially inflating this measurement of mobile subscribers, this is unlikely to be the case in African nations. Considering the dire poverty most African nations face, it is highly unlikely that a large number of the individuals living in these countries will find it economically feasible or even possible to own multiple mobile phone subscriptions. 5. Summary statistics for each of the variables used in this analysis are included in the relevant regression tables. Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Corruption in Africa 351 6. This measure is derived from the World Bank’s Voice and Accountability governance indicator. This measure represents the percentile rank of the strength of democratic practices available in that nation. For more information, please refer to http://www.worldbank.org/wbi/governance/ index.htm. 7. For more information regarding how the World Bank constructs the Regulatory Quality Index, please visit http://www.worldbank.org/wbi/governance/index.htm. 8. The World Bank does not provide measures for the Voice and Accountability and Regulatory Quality indexes for the years 1999 and 2001. To address this, I calculate the average of the years immediately preceding and following 1999 and 2001. For example, the 1999 value included in this analysis is the average of each country’s 1998 and 2000 scores. 9. For a detailed timeline of anti-corruption activities in Namibia, please refer to http:// www.anticorruption.info/ corr_in_time.php. 10. Since the Afrobarometer survey was conducted in the first months of 2006 (February 14 to March 7), it is reasonable to assume that the 2005 map of mobile signal coverage in Namibia is representative of the degree of coverage present in early 2006 when the respondents were surveyed. 11. One could argue that population density is correlated with corruption. However, a Pearson product-moment correlation of the percentage of a nation’s population that lives in an urban area and each nation’s CPI reveals that these factors are uncorrelated (−.06). In regard to a possible correlation between accessibility and corruption, the most obvious determinant of accessibility is the roadway infrastructure extant in a province. Essentially, the more roads, the easier it will be for mobile service providers to erect cell towers in a particular region. As to whether the degree of roadway infrastructure is correlated with corruption levels at the national level, roadway density [total roadways (km)/total land area (km2)] is uncorrelated with perceived corruption (Pearson productmoment correlation: .1). At the provincial level in Namibia, road density is only very weakly correlated with the average level of corruption that residents assigned to the following in 1999: local government corruption (−.21 ), presidential corruption (−.22), and likelihood of paying bribe (−.33). Moreover, the relationship that these three average perceptions of corruption share with road density fails to reach statistical significance in all three cases. 12. This map was published in the Information Society and Development Conference’s The Acacia Atlas 2005: Mapping African ICT Growth, based on information provided by the GSM Association. For more details and to view the map, please visit http://www.idrc.ca/uploads/user-S/ 11836495021Acacia_Atlas_ 2005.pdf. References Abraham, R. (2007). Mobile phones and economic development: Evidence from the fishing industry in India. Information Technologies and International Development, 4(1), 5–17. Aker, J. C. (2008). Does digital divide or provide? The impact of cell phones on grain markets in Niger. Unpublished doctoral dissertation, University of California–Berkeley. Allison, P. D. (2005). Fixed effects regression methods for longitudinal data using SAS. Cary, NC: SAS Institute. Azfar, O., Lee, Y., & Swamy, A. (2001). The causes and consequences of corruption. Annals of the American Academy of Political and Social Science, 573, 42–56. Bardhan, P., & Mookherjee, D. (2000). Capture and governance at local and national levels. American Economic Review, 90, 135–139. Best, M. L., & Wade, K. W. (2005). The Internet and democracy: Global catalyst or democratic dud? Retrieved from http://cyber.law.harvard.edu/publications Buys, P., Dasgupta, S., Thomas, T., & Wheeler, D. (2008). Determinants of a digital divide in SubSaharan Africa: A spatial-economteric analysis of cell phone coverage. (The World Bank policy research working paper no. WPS 4516). Washington, DC: The World Bank Sustainable Rural and Urban Development Team, Development Research Group. Daly, J. A. (2000). Studying the impacts of the Internet without assuming technological determinism. Aslib Proceedings, 52, 285–300. Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 352 Catie Snow Bailard Das, J., Dercon, S., Habyarimana, J., & Krishnan, P. (2002). Rules vs. discretion: Public and private funding in Zambian basic education. Part 1: Funding equity. Washington, DC: World Bank. Donner, J. (2005). The use of mobile phones by microentrepreneurs in Kigali, Rwanda. Changes to Social and Business Networks Information Technology and International Development, 3(2), 3–19. Frost & Sullivan. (2006). Social impact of mobile telephony in Latin America. Retrieved from the GSM Latin America Association Web site: http://www.gsmlaa.org/files/content/0/94/ Social%20Impact%20of%20Mobile%20Telephony%20in%20Latin%20America.pdf Greenstone, J. D. (1966). Corruption and self-interest in Kampala and Nairobi. Comparative Studies in Society and History, 8, 199–210. Harsch, E. (1993). Accumulators and democrats: Challenging state corruption in Africa. Journal of Modern African Studies, 31, 31–48. Huntington, S. P. (1968). Political order in changing societies. New Haven, CT: Yale University Press. International Telecommunications Union. (2008). African Telecommunication/ICT indicators 2008: At a crossroads. Geneva, Switzerland: Author. Itano, N. (2005). Africa’s cell phone boom creates a base for low-cost banking. Christian Science Monitor. Retrieved from http://www.csmonitor.com/2005/0826/p07s01-woaf.html Jackson, R. H., & Rosberg, C. G. (1994). The political economy of African personal rule. In D. Apter & and C. Rosberg (Eds.), Political development and the new realism in sub-Saharan Africa (pp. 291–322). Charlottesville, VA: University of Virginia Press. Jensen, R. (2007). The digital provide: Information (technology), market performance and welfare in the South Indian fisheries. Quarterly Journal of Economics, 3, 879–924 Kedzie, C. R., with Aragon, J. (1997). Coincident revolutions and the emergent dictator’s dilemma: Thoughts on communication and democratization. In J. E. Allison (Ed.), Technology, development, and democracy: International conflict and cooperation in the information age (pp. 105–130). Albany, NY: State University of New York Press. Klitgaard, R. (1988). Controlling corruption. Berkeley, CA: University of California Press. LaFraniere, S. (2005). Cellphones catapult rural Africa to 21st century. New York Times. Retrieved from http://www.nytimes.com/2005/08/25/international/africa/25africa.html?ex=1282622400& en=32b49363eac57aae&ei=5090&partner=rssuserland&emc=rss Lane, B., Sweet, S., Lewin, D., Sephton, J., & Petini, I. (2006). The economic and social benefits of mobile services in Bangladesh: A case study for the GSM Association. Retrieved from http:// www.dirsi.net/english/files/Ovum%20Bangladesh%20Main%20report1f.pdf Leff, N. (1964). Economic development through bureaucratic corruption. American Behavioral Scientist, 8(3), 8–14. Leonard, D. K., & Straus, S. (2003). Africa’s stalled development: International causes and cures. Boulder, CO: Lynne Rienner. LeVine, V. T. (1975). Political corruption: The Ghana case. Stanford, CA: Hoover Institution Press. Mauro, P. (1995). Corruption and growth. Quarterly Journal of Economics, 110, 681–712. Mauro, P. (1997). The effects of corruption on growth, investment, and government expenditure: A cross-country analysis. In K. A. Elliott (Ed.), Corruption in the global economy (pp. 83–107). Washington, DC: Institute for International Economics. Molony, T. (2005). Food, carvings and shelter: The adoption and appropriation of information and communication technologies in Tanzanian micro and small enterprises. Unpublished doctoral dissertation, University of Edinburgh. Montinola, G. R., & Jackman, R. W. (2002). Sources of corruption: A cross-country study. British Journal of Political Science, 32, 147–170. Ngowi, R. (2005). Africa’s cellphone explosion changes economics, society. USA Today. Retrieved from http://www.usatoday.com/tech/products/gear/2005-10-16-africa-cellular_x.htm Ogendeji, O. A. (2008). GSM operator supports fight against corruption. Network World. Retrieved from http://www.networkworld.com/news/2008/061608-gsm-operator-supports-fight-against.html Downloaded By: [Bailard, Catie Snow] At: 22:45 4 August 2009 Corruption in Africa 353 Olivier de Sardan, J. P. (1999). A moral economy of corruption? Journal of Modern African Studies, 37, 25–52. Platteau, J.-P., & Gaspart, F. (2003). The risk of resource misappropriation in community-driven development. World Development, 31, 1687–1703. Price Waterhouse Coopers. (1999). Health and Education Financial Tracking Study. Report prepared for government of Tanzania. Rasmusen, E., & Ramseyer, J. M. (1994). Cheap bribes and the corruption ban: A coordination game among rational legislators. Public Choice, 78, 305–327. Rautray, S. (2006). Indian politician’s son guilty of model’s murder. Free Republic. Retrieved from http://www.freerepublic.com/focus/f-news/1755260/posts Reinikka, R., & Svensson, J. (2004). Local capture evidence from a central government transfer program in Uganda. Quarterly Journal of Economics, 119, 679–706. Reinikka, R., & Svensson, J. (2005). Fighting corruption to improve schooling: Evidence from a newspaper campaign in Uganda. Journal of the European Economic Association, 2, 259–267. Rose-Ackerman, S. (1993). Corruption: A study in political economy. New York: Academic Press. Ross, W. (2004). Mobile markets deny middlemen. BBC News. Retrieved from http://news. bbc.co.uk/go/em/fr/-/2/hi/africa/3321167.stm Samuel, J., Shah, N., & Hadingham, W. (2005). Mobile communications in South Africa, Tanzania and Egypt: Results from community and business surveys. Vodafone Policy Paper Series, 2, 44–52. Sandholtz, W., & Koetzle, W. (2000). Accounting for corruption: Economic structure, democracy, and trade. International Studies Quarterly, 44, 31–50. Schleifer, A., & Vishny, R. W. (1993). Corruption. Quarterly Journal of Economics, 108, 599–617. Sullivan, K. (2006, July 9). In war-torn Congo, going wireless to reach home: For poor, cellphones bridge digital divide. Washington Post, p. A1. Szetfel, M. (2000). Clientelism, corruption, & catastrophe. Review of African Political Economy, 27, 427–441. Tanzi, V. (1998). Corruption around the world: Causes, consequences, scope, and cures. Staff Papers of the International Monetary Fund, 45, 559–594. Treisman, D. (2000). The causes of corruption: A cross-national study. Journal of Public Economics, 76, 399–457. Ye, X., & Canagarajah, S. (2002). Efficiency of public expenditure distribution and beyond: A report on Ghana’s 2000 Public Expenditure Tracking Survey of Primary Health and Education. Washington DC: World Bank.