Has Centralization Increased Corruption in Thailand?
by Joel Selway
Picture: Thai baht
Source: Flickr (Peter Hellberg)
Over the past month there has been intense discussion concerning the role of decentralization in Thailand’s future political structure. In July, the military government ordered a halt to direct elections of mayors and local councils (link). Part of the justification was the high levels of corruption reported in local agencies. However, Kriangkrai Phoomlaochaeng, a member of the National Reform Council and chairman of the National Municipal League of Thailand, blamed corruption on a lack of funds, calling for an increase in the budget for Elected Local Administration Organisations (LAOs) to 35% (from 27%) of the annual state budget (link). But what do the data say? Does decentralization increase or reduce corruption? Does it depend on how much money is channeled to local councils?
First, let’s turn to Thailand’s neighbors in Southeast Asia. A recent study of decentralization in Thailand’s neighbor, Vietnam, found that local councils enabled citizens to rein in local corruption (here). Provinces that were given higher transfers without local councils to check leaders’ power, however, saw no improvement in corruption (Malesky et al 2014). Pepinsky and Wihardja (2011) find the opposite in Indonesia about local councils. The 1999 decentralization laws increased corruption due to the creation of “little kings who are unaccountable to their citizens . . . and no longer constrained in their confiscatory impulses by a strong central government” (p. 361). The authors blame this on money politics and business elites funding local campaigns. Such studies should serve as a guide for a similar study of decentralization in Thailand. They are exemplary in that they both marshal immense amounts of data, such as individual surveys and provincial-level statistical data. To date, I have seen no comparable study of decentralization and corruption in Thailand. Most of the studies I have read are of a qualitative nature.
QUALITATIVE MEASURES OF DECENTRALIZATION
A simple first step I thus take in this post is to rely on the qualitative expertise of Dufhueset al. (2011) and simply correlate their assessment of how decentralized Thailand has been over time to annual measures of corruption. Figure 1 is taken directly from their paper. It shows a rapid increase in decentralization in the 1990’s culminating in the passing of the Decentralization Act in 1999. Interestingly, the trend then flattens, showing no increase in decentralization after the act was passed. The authors blame this on Thaksin’s “recentralization” that countered the written legal code (see also Mutebi 2004). Finally, decentralization weakens even further after the 2006 coup. As Dufhues et al. state: “However, since the coup in 2006, laws and regulations have been passed to tighten the connection between central and local government. For instance, the Provincial Administration Act was revised and as a result, the role of bureaucrats, especially those from the Ministry of Interior was strenghtened at the local level (Chardchawarn 2010) . . . Furthermore, the new 2007 Constitution effectively reduced the power of political parties and strengthened the hand of unelected bureaucrats (Wong 2007)."
I have overlaid Dufhues et al.’s graphic with Transparency International’s Corruption Perceptions Index, for which there is data between 1995 and 2011. The thinner red line is the actual score with each point denoting a separate year; the thicker red line is the smoothed function. Higher scores actually represent more transparency (so less corruption). The pattern suggests that changes in corruption roughly track decentralization, albeit with a four-year lag. Decentralization reaches its peak in 2001, but transparency not until 2005. This is partially explained by the yearly CPI score relying on surveys from the previous two years as explained in the Transparency International’s methodology (here). I want to emphasize that the change in corruption is still only moderate. CPI scores range theoretically from 0 to 10, and Thailand’s range is a mere 0.8 in the 16 years of the index.
There seems to be something here, but Dufhues et al.’s estimation of decentralization is a/. not particularly systematic and b/. gives us no idea of how Thai decentralization compares globally. Shair-Rosenfieldet al.(2014) solve both these problems with their new Regional Authority Index, which provides “annual estimates across ten dimensions of regional authority” for five Asian countries across six decades. The richness of their measure is immense, but unlike Dufhues et al. it only captures formal rules. Nevertheless, it provides us a second test of the hypothesis on decentralization and corruption. In Figure 2, a truncated version of the original graphic, we see that formal decentralization actually increased continually since the Decentralization Act of 1999. Plotting the CPI scores on this measure tells a similar story to Dufhues et al. up until the 2006 coup. Then, despite increases in formal decentralization, corruption gets worse.
The political science literature more broadly has made additional progress in unpacking the several components of decentralization and conceptualizing and measuring those components in a way that allows for meaningful comparison across countries. The first major distinction in the literature is between political and fiscal decentralization. Indeed, it is this distinction that Kriangkrai seems to be suggesting in his call for more funds to LAOs: fiscal decentralization in Thailand has not matched political decentralization. Is he right?
The most common way to measure fiscal decentralization is the proportion of subnational governments’ budgets, either of the total government budget or of the gdp. Using this measure, scholars have mixed findings (Treisman 2007). Where does Thailand stand on this measure compared to other countries? Table 1 shows Thailand in comparison to a number of its regional neighbors as well as the countries at the first, fiftieth, and ninety-ninth percentiles. We see that Thailand falls a little below the median within Asia in terms of subnational expenditures as a percentage of the overall government budget. Moreover, there are countries with higher subnational expenditure shares with the same or worse corruption scores. The two countries with much better corruption scores are notably much wealthier (Japan and Korea). Most countries with lower fiscal decentralization have worse corruption scores but these are all also notably poorer countries than Thailand. The exception again is Malaysia, which also happens to be wealthier.
If the recent NCPO decision to cut LAO budgets in half is implemented (link), Thailand would fall to somewhere between Bangladesh and Nepal. But would this mean that its corruption would fall too? That is much less clear.
The same pattern holds when looking at the share of government revenues brought in by subnational units, a second measure of fiscal decentralization. Indonesia and Pakistan drop beneath Thailand now, which produces a slightly stronger correlation with corruption. Even taking Japan and Korea out of the equation, the average CPI score for countries with higher revenues than Thailand is 3.23 compared to 2.38 for countries lower than Thailand (including Malaysia it is 2.67). But again, controlling for gdp per capita this relationship would almost certainly disappear. In short, there is not much to learn from this crude assessment, which seems to support the decidedly mixed findings on fiscal decentralization.
Comparing across regions doesn’t provide any additional insight. Thailand and Brazil have similar expenditures and revenues and a similar CPI score. We might also put South Africa in the same grouping. Costa Rica is low on both, but like Malaysia has a higher CPI score. Both Gabon and Armenia have low expenditure and revenue shares and accompanying low CPI scores.
Next, I see if changes in fiscal decentralization over time within Thailand correlates with its corruption score. Figure 2 shows the subnational share of total government expenditures between 1998 and 2011 (blue line). Data on expenditures are from Kosiyanon (2012). We can see that this sharply increased following the implementation of the 1999 Decentralization Act in 2001 and has continued to steadily increase over the last decade. There is about a four-year lag again before the improvement in Thailand’s corruption score, but then there is an inexplicable drop from which it currently has not recovered that the continual increase in fiscal decentralization cannot explain. In sum, there seems to be a weak correlation within Thailand.
A third measure of fiscal decentralization is the subnational government employment share. Treisman (2008) is the only study I know that does this and he finds that countries with a larger number of local government employees experience more bribery. Thailand ranks 36th out of 91 countries in Treisman’s dataset, just outside the top 1/3 of countries, with 41% of all government employees being at the local level. For this set of countries, the share of government employees does seems to be fairly well correlated with the other fiscal decentralizaiton measures. Countries with lower shares (Indonesia and Philippines) have lower CPI scores, but those with higher shares are not convincingly better, especially when we exclude Korea and Japan.
Looking more broadly across the world, we see that countries with very high subnational shares of total government employment (>70%) seem to share China and Vietnam’s low CPI scores—even wealthier Argentina. Low shares also seem to result in low CPI scores, as within the Asia sample. This perhaps indicates that medium proportions (40-70) are the best—too high a share and there are too many opportunities for local corruption, whereas too low a share means too many opportunities for corruption at the center. This is a phenomenon worth investigating in more detail, but it would seem that Thailand lies within the moderate range.
Lastly, I turn to political decentralization. This is a much trickier concept to measure, mainly because of the numerous forms decentralization can take. The most prominent study on decentralization, by DanielTreismanet al. (2007), categorizes political decentralization along several features. I will discuss just two here. The first feature is the number of administrative tiers. This may affect corruption because the more complex is decentralization, the harder it is to keep local leaders accountable. Treisman et al. find strong evidence that this increases corruption. A second feature is the bottom unit size, which measures the surface area of the administrative districts. The connection to corruption is that the more spread out an administrative unit is the easier it is to hide corrupt practices. However, they find no crossnational evidence that this second feature affects corruption.
Where does Thailand fall comparatively on these measures?
Countries with larger unit sizes in Asia (Malaysia, Brazil and S. Africa) do seem to have lower levels of corruption. However, most Asian countries have very small unit sizes compared to the rest of the world, which could explain the low levels of CPI in this region. The world average is 2.05. The average of this sample is 0.27.
Small unit size is somewhat related to the number of tiers, although the number of tiers is more about the administrative complexity than the physical ability to hide corruption. This was the feature with the strongest empirical support in Treisman’s study. For the sample here, small tier size connects Japan, Malaysia, and South Africa, with Armenia being an exception. The average CPI score for countries with three tiers is 4.26; for four tiers the average is 3.52; and for the rest it is 3.
This high number of tiers results in small (population-wise rather than area-wise) units and has been noted in qualitative accounts of Thailand’s decentralization. As Dufhues et al. state: “the average Tambon cannot support a high school or professional administrative staff . . . By creating huge numbers of small units, the superiority of the central agents is always maintained.”
In concluding, I note one final study of interest. Lessmannand Markwardt(2009) argues that decentralization can reduce corruption, but only where there is a strong free press. According to Freedom House’s Freedom of the Press index, Thailand has grown progressively worse in this area since 2002 when it was awarded a 30/100 (lower numbers are better), a score similar to Italy, and labeled as “Free”. In 2003, as a result of Thaksin’s increased pressure on media outlets, Thailand dropped to the “Partly Free” category, and has been on a downward spiral ever since. In 2011, it dropped to the “Not Free” category. This does not bear well on the future of decentralization and corruption in Thailand.
 In terms of its relative placement the changes are slightly more significant: while Thailand would have moved up only 7 places (from 80 to 73 out of 183) if it had scored its high of 3.8 in 2011 (rather than its actual score of 3.4) if it had scored its low of 3.0, it would have moved down 20 places to 100th.
 This is mitigated by local councils having the ability to raise revenue themselves; where they are allowed, the effect of the number of tiers is reduced.
 Third, are they elected? Fourth, do they have autonomous powers, i.e. constitutional authority to legislate in certain specified that are not explicitly subject to central laws. Fifth, do they have residual powers, or the ability to legislate on areas not explicitly assigned to other levels?