TY - JOUR
AU - Lin,, Faqin
AB - Abstract This paper examines effects of exogenous income from international commodity price windfalls on HIV infections in a panel of sub-Saharan African countries during the period 1985–2007. The main finding is that an increase in income leads to a significant rise in HIV infections in autocratic countries while there is no significant effect in democracies. Further analysis suggests that increasing urbanisation and decreasing public health expenditure share in GDP in autocracies are the dominant channels behind such distinct comparison. After controlling for urbanisation and public health expenditure share, the effect of income on HIV infection rates shrinks drastically and is statistically insignificant. 1. Introduction The HIV has infected nearly 1 in every 20 adults in the sub-Saharan African (SSA) countries, which makes up around 70% of HIV patients worldwide1; meanwhile, SSA countries are among the poorest in the world and produce only about 1% of world output with almost 10% of its population in 2010 (WDI, 2014). This extreme combination provides a meaningful and content-rich background to test the impact of income on health, whether it follows the ‘wealthier is healthier’ principle in Pritchett and Summers (1996), i.e., increasing income helps improve health, or it complies more with the more recent differing opinions as in Case and Deaton (2005), Ruhm (2007), and Oster (2012). From a theoretical point of view, the effect of a change in income on the HIV infection rate is ambiguous. The traditional ‘wealthier is healthier’ view rests on a simple but plausible logic, i.e., higher incomes allow people to increase health-related spending, both public and private, and reduce unhealthy behaviour, which is also applicable for analysis on HIV infection in SSA countries (e.g., Burke et al., 2015). One very important channel of HIV infection in SSA countries is through transactional sex, which could be broadly interpreted as sex in exchange for money, goods, or services in the form of prostitution, ‘dating’ or even marriage (Robinson and Yeh, 2011). A variety of researches (Clark, 2004; Swidler and Watkins, 2007; Dinkelman et al., 2008; Robinson and Yeh, 2011; LoPiccalo et al., 2012) has documented that the supply of transactional sex rise when faced with economic hardship. Therefore, it is not surprising that cash transfer programmes could potentially help to control the spread of HIV (Baird et al., 2012; de Walque et al., 2012). However, this is only part of the story. If transitional sex is a normal or luxury good, increase in income could also escalate its demand of sex and even the demand of risky sex activities (Ahlburg and Jensen, 1998; Kohler and Thornton, 2012). What makes things more complicated is migration. As argued in Ahlburg and Jensen (1998), migrants increase their demand for commercial sex because of the absence of wives and absence of family monitoring, etc., which suggests that increasing migration helps spread HIV (Lurie et al., 2003). However, people migrate in front of negative economic shocks in search of work (Skoufias, 2003), but people also migrate when an economy is prosperous and involves in more trade (Oster, 2012; Lin and Sim, 2013, 2015), which makes the effect of income on HIV infection through the migration channel really uncertain. To fully investigate this problem, we use the variation in an international commodity price as an instrument for national income to prevent from potential reverse causality and omitted variable problems, and exploit the panel structure in our dataset based on 48 SSA countries during 1985–2007 that helps purge both time invariant cross-country heterogeneity and SSA wide shocks. Our instrumental variable for national income is well established in the literature (e.g., Deaton, 1999; Brueckner and Ciccone, 2010; Arezki and Brueckner, 2012a, b). Particularly, a key characteristic of the SSA countries that makes this estimation strategy plausible is that these countries are highly dependent on commodity exporting sector. Hence variations in international commodity prices can induce substantial variation in real per capita GDP growth through changes in terms of trade. Because the economic size of SSA countries (as measured by the share in world commodity production) is extremely small (so that the country can be effectively treated as being a price taker on the international commodity market) the induced variations in per capita GDP growth will be exogenous to variations in economic growth. Oster (2012) documents that international goods trade could hasten the spread of HIV, and therefore in our analysis we also control for trade openness in our regression to address this concern. We first estimate the average effect that income from international commodity price windfalls has on HIV infections. We find that international commodity price shocks do not show significant influence on HIV infection. However, a close scrutiny of the results suggests that while the impact of income on HIV infection is almost neutral in democracies, it is quantitatively large and statistically significant in autocracies, making it non-negligible. More specifically, a 1% increase in income is associated with an about 0.8% increase in the rate of HIV infections in autocracies; in democracies the effect is quantitatively small and statistically indistinguishable from zero. In addition, we can reject the hypothesis that the effect of income on HIV infections is the same in democracies and autocracies. These results are not really unexpected: in autocracies people cannot punish their governments through the electoral process if unhealthy living conditions persist (Acemoglu et al., 2004; Padró i Miquel, 2007; Besley and Kudamatsu, 2008). Evidence that health condition is worse in autocracies can be found, for example, in Besley and Kudamatsu (2008) and Kudamatsu (2012). In order to fully understand why income from international commodity price windfalls increases the rate of HIV infections in autocratic SSA countries, we investigate channels through which income could affect HIV infections. One important channel is urbanisation. As noted in, for example, Bolnik et al. (2006), HIV prevalence in SSA countries is significantly higher in urban areas.2 If preventative measures are not taken, the virus spreads more rapidly in urban areas than in rural areas because of lower costs of finding sexual partners in cities. Meanwhile, at the country level higher incomes are known to be associated with urbanisation (Brueckner, 1990, 2012; Henderson, 2003). As people move from the countryside to the city, where costs for human interaction are lower, governments are faced with significant public management challenges. Autocracies tend to fare worse than democracies in terms of optimally managing the size of cities, and Davis and Henderson (2003) actually document that autocracies promote excessive concentration. Glaeser (2014) argues that in poor countries, autocratic governments are faced with severe challenges of implementing policies that mitigate negative externalities from urbanisation. In addition, urbanisation usually leads to migration (Todaro, 1997; Potts, 2009), which also helps spread HIV. We find that in autocracies, income from international commodity price windfalls leads to a significant increase in the urbanisation rate. After controlling for urbanisation, the effect of income from international commodity price windfalls on HIV infections decreases by about one-third. This suggests that urbanisation is an important channel through which income affects HIV incidence. The increase in the demand for sex that arises from higher incomes does not have to lead to higher HIV incidence, if precautions are taken. Governments can reduce the spread of HIV in cities by providing public health services.3 Indeed, our panel regressions show that HIV infections are significantly negatively correlated with public health expenditures as share of GDP. We do not, however, find any evidence that in autocracies income from international commodity price windfalls is used to increase public health expenditure shares. On the contrary, the panel estimates show that income from international commodity price windfalls has a significant negative effect on public health expenditures in autocratic SSA countries. The negative effect of income on public health expenditures is consistent with the view that autocracies tend to provide worse public health services than democracies (Kaufman and Segura-Ubiergo, 2001; Tavares and Wacziarg, 2001; Brown and Hunter, 2004; Avelino et al., 2005; Stasavage, 2005; McGuire, 2006; Acemoglu and Robinson, 2012). The fact that international commodity price windfalls actually lead to less public health expenditure is also consistent with the ‘voracity effect’ (Tornell and Lane, 1999), which states that a positive shock could perversely create a more-than-proportional increase in fiscal redistribution and thus hurts economic growth. Conditional on urbanisation and public health expenditures, the instrumental variables estimates show that the effect of income on HIV infections is statistically insignificant. Quantitatively, this effect (that, by construction, must go through channels other than urbanisation and public health expenditures) is less than one-tenth the size of the unconditional effect. For comparison, when controlling for urbanisation only, the conditional effect of income on HIV infections is about two-thirds the size of the unconditional effect. This suggests that urbanisation and the disappearance of public health expenditures are important channels through which an exogenous increase in income raises HIV infections in autocratic SSA African countries. The result is consistent with the view in the literature (e.g., Oster, 2005), that the high incidence of HIV in the SSA region is due to high transmission rates, but the spread of HIV may be prevented by providing readily available, off-patent drugs and such drugs can be more effectively distributed (by public health services) in cities than in the countryside. Our paper contributes to literature that investigates how income affects health. Similar to Pritchett and Summers’ (1996) ‘wealthier is happier’ view, Levine and Rothman (2006) document that trade can benefit the health of children through the income channel. On the other hand, in line with Case and Deaton (2005), Biggs et al. (2010) argue that higher GDP per capita has no significant effect on health measures, and Rajan et al. (2013) caution that rising income levels have not been matched by improvements in public health in India. With regard to HIV in particular, Ahlburg and Jensen (1998) point out that the increased income provided by exports could drive more risky sex, which may drive up HIV infection rates, and Oster (2012) documents that increased international goods trade in SSA Africa, thanks to risky sex practices of truck drivers, is associated with increases in HIV. Our paper contributes to that literature by (i) using plausibly exogenous variation in income that is driven by variations in the international commodity prices; (ii) examining heterogeneity between democracies and autocracies; and (iii) investigating channels through which international commodity price windfalls affect the relationship between income and HIV. The remainder of the paper is structured as follows. The data are presented in Section 2. Section 3 explains the estimation methodology. Section 4 presents the estimation results of the impact that income has on HIV infections. Section 5 presents estimation results on the channels through which income may affect the rate of HIV infections. Section 6 concludes. 2. Data Our dataset consists of a panel of 48 SSA countries during the period 1985–2007. The time span and coverage of countries are determined by the availability of data. Our panel dataset is constructed from the following main sources: Arezki and Brueckner (2012a, b), Penn World Tables, the World Development Indicators, and Oster (2012). The key variables in our study are described below. 2.1 Commodity price windfalls Given that SSA countries are predominantly primary commodity exporters, the global demand for commodities could have strong influence on their national incomes. If the export of a certain commodity is important to a country, the global demand of that commodity (reflected by its international price) would have a greater impact on the income of that SSA country. This allows us to investigate the country-specific effect of international commodity prices on income, where the country specificity of this effect is based on differences in the importance of various exported commodities across different SSA countries. Data on international commodity net-export price index are from Arezki and Brueckner (2012a, b). The index is constructed as ComPIi,t=∏cComPriceC,tθi,C(1) where log(ComPriceC,t) is the international price of commodity c in year t ⁠, and θi,C is the average (time invariant) value of net-exports of commodity c in the GDP of country i ⁠. We obtain data on annual international commodity prices for the 1985–2007 period as well as data on the value of commodity exports from UNCTAD Commodity Statistics.4 The commodities included in our index are aluminium, bananas, beef, cocoa, coffee, copper, cotton, gold, iron, maize, lead, oil, pepper, rice, rubber, sugar, tea, tobacco, wheat, wood and zinc. In the case where there were multiple prices listed for the same commodity, we use the simple average of all the relevant prices. 2.2 Democracy Democracy is measured by the revised combined Polity score (Polity2) of the Polity IV database (Marshall and Jaggers, 2009). The Polity2 score ranges from –10 to + 10. A score of 10 reflects the most democratic institution; a score of –10 reflects the most autocratic institution, and a score of zero indicates a political institution that is neither democratic nor autocratic.5 In the literature, the Polity2 score has been used to distinguish democracies from autocracies. One example is Arezki and Brueckner (2012b), who examine whether the effect of international commodity price windfalls on external debt is contingent on whether countries are democratic or autocratic. They code democratic (autocratic) institutions as strictly positive (negative) values of the Polity2 score and run separate regressions of external debt on international commodity price windfalls for democracies and autocracies (see, also, Brueckner et al., 2012). In this paper, we identify democracies and autocracies in the same way as Arezki and Brueckner (2012b). As a robustness check, we will employ the democracy indicator of Przeworski et al. (2000). 2.3 HIV infections and income The HIV infection rate for period t is the rate of new infections in that period divided by the total population (times 100, so that the value is in percent). The data are from Oster (2012). Oster uses the following formula for computing the HIV infection (incidence) rate: πit=hit−hi,t−1+dit ⁠, where hit is HIV prevalence (in 100 population) and dit denotes the number of deaths (in 100 population) in the current year. Oster (2012) uses UNAIDS data on estimates of HIV prevalence and tracks back to the start of the epidemic (a combination of earlier trend data from UNAIDS and linear interpolation technique). She uses the following equation πit=hit−hi,t−1+∑j=1t−1ct−jπj to calculate incidence, and denote the chance of dying from HIV after τ years of infection as cτ which is estimated based on other sources of data.6 As a robustness check, we will draw on mortality-based HIV infection rates computed in Oster (2012). Oster uses the following formula dit=c1πit−1+c2πit−2+⋯+c20πit−20 for computing the infection rate πit ⁠, where dit is the number of deaths from HIV in a given year, and cτ is the chance of dying from HIV after τ years of infection. She also needs to see data back to the start of the epidemic and other data sources for the calculation (e.g., death rates and time to death) as the computation for UNAIDS HIV infection rates. However, One shortcoming of the mortality-based data for our purposes of analysis is that it only provides a limited number of observations (about one-fifth of the UNAIDS data). We therefore use Oster's UNAIDS data as our main measure of HIV infections. Our main source of data on national income, measured by real GDP per capita data, is the Penn World Tables, version 6.3 (Heston et al., 2009). As a robustness check, we use data on real GDP per capita from WDI (2014).7 2.4 Other variables Data on trade openness (exports plus imports divided by GDP) are from NBER–United Nations Trade Data. Data on the urbanisation rate (urban population as share of total population) and public health expenditures (share of GDP) are from WDI (2014). Summary statistics of the main variables used in the econometric analysis are presented in Table 1. Table 1: Summary Statistics Variable . Mean . Std. dev. . Min . Max . HIV incidence rate 0.745 0.950 0.000 5.859 Lagged HIV prevalence 4.721 6.342 0.000 28.900 HIV incidence rate, mortality based 0.589 0.558 0.000 2.486 Log (GDP p.c.), PWT 7.241 0.808 4.777 10.062 Log (GDP p.c.), WDI 6.357 1.042 3.913 9.519 Polity2 −1.178 5.770 −10.000 10.000 Przeworski et al. (2000) democracy 0.303 0.460 0.000 1.000 Trade openness (share of GDP) 0.609 0.585 0.059 9.866 Urbanisation rate 0.319 0.143 0.051 0.844 Public health expenditures (share of GDP) 0.080 0.099 0.035 0.205 Variable . Mean . Std. dev. . Min . Max . HIV incidence rate 0.745 0.950 0.000 5.859 Lagged HIV prevalence 4.721 6.342 0.000 28.900 HIV incidence rate, mortality based 0.589 0.558 0.000 2.486 Log (GDP p.c.), PWT 7.241 0.808 4.777 10.062 Log (GDP p.c.), WDI 6.357 1.042 3.913 9.519 Polity2 −1.178 5.770 −10.000 10.000 Przeworski et al. (2000) democracy 0.303 0.460 0.000 1.000 Trade openness (share of GDP) 0.609 0.585 0.059 9.866 Urbanisation rate 0.319 0.143 0.051 0.844 Public health expenditures (share of GDP) 0.080 0.099 0.035 0.205 Table 1: Summary Statistics Variable . Mean . Std. dev. . Min . Max . HIV incidence rate 0.745 0.950 0.000 5.859 Lagged HIV prevalence 4.721 6.342 0.000 28.900 HIV incidence rate, mortality based 0.589 0.558 0.000 2.486 Log (GDP p.c.), PWT 7.241 0.808 4.777 10.062 Log (GDP p.c.), WDI 6.357 1.042 3.913 9.519 Polity2 −1.178 5.770 −10.000 10.000 Przeworski et al. (2000) democracy 0.303 0.460 0.000 1.000 Trade openness (share of GDP) 0.609 0.585 0.059 9.866 Urbanisation rate 0.319 0.143 0.051 0.844 Public health expenditures (share of GDP) 0.080 0.099 0.035 0.205 Variable . Mean . Std. dev. . Min . Max . HIV incidence rate 0.745 0.950 0.000 5.859 Lagged HIV prevalence 4.721 6.342 0.000 28.900 HIV incidence rate, mortality based 0.589 0.558 0.000 2.486 Log (GDP p.c.), PWT 7.241 0.808 4.777 10.062 Log (GDP p.c.), WDI 6.357 1.042 3.913 9.519 Polity2 −1.178 5.770 −10.000 10.000 Przeworski et al. (2000) democracy 0.303 0.460 0.000 1.000 Trade openness (share of GDP) 0.609 0.585 0.059 9.866 Urbanisation rate 0.319 0.143 0.051 0.844 Public health expenditures (share of GDP) 0.080 0.099 0.035 0.205 3. Methodology Our main estimating equation relates the HIV infection rate (⁠ HIVi,t ⁠) to the log of real income per capita log(GDP per capitai,t) for country i at year t as HIVi,t=βlog(GDP per capitai,t)+∅’Xi,t+μi+μt+vi,t,(2) where Xi,t is the set of control variables including trade openness, motivated by Oster (2012), and the lagged HIV prevalence rate. Our objective is to estimate the causal effect of income on HIV incidence, which is summarised by β ⁠. We control for country fixed effects, μi ⁠, and year fixed effects, μt ⁠. Fixed effects estimation is a powerful way of soaking up the determinants that we do not observe or cannot easily control for.8 However, it does not solve the issue of reverse causality. To address this, we use commodity price windfalls as an instrument to identify plausibly exogenous variation in SSA countries’ national income. The first stage of our instrumental variables analysis relates the log of GDP per capita to the log of the international commodity net-export price index: log(GDP per capitai,t)=γlog(ComPIi,t)+δ’Xi,t+ai+at+ui,t(3) Equation (2) is estimated using two-stage least squares. We also estimate the effect of international commodity price windfalls on HIV infections by looking at the reduced form equation: HIVi,t=ρlog(ComPIi,t)+φ’Xi,t+bi+bt+ei,t(4) 4. Empirical results 4.1 Estimation results Table 2 reports OLS estimates of the relationship between income and the HIV infection rate. While the OLS regression is not equipped to deal with the issue of reverse causality, and thus its estimates do not identify a causal effect, it is nonetheless useful for helping us understand the consequences of omitting fixed country characteristics and SSA Africa wide shocks. Column (1) reports estimates from a simple bivariate regression. Without controlling for time and country fixed effects, the OLS regression suggests that income and HIV are positively correlated. This relationship still holds after including lagged HIV and trade openness, see columns (2) and (3). Table 2: OLS Estimates of the Relationship Between Income and HIV . (1) . (2) . (3) . (4) . (5) . Dependent variable: HIV incidence rate  Log (GDP p.c.) 0.424*** 0.120** 0.121** 0.011 0.212 (0.142) (0.055) (0.056) (0.074) (0.145)  Lagged HIV prevalence No Yes Yes Yes Yes  Trade openness No No Yes Yes Yes  Country effect No No No Yes Yes  Year effect No No No No Yes  Observations 828 828 828 828 828  Adj. R2 0.130 0.572 0.573 0.664 0.706 . (1) . (2) . (3) . (4) . (5) . Dependent variable: HIV incidence rate  Log (GDP p.c.) 0.424*** 0.120** 0.121** 0.011 0.212 (0.142) (0.055) (0.056) (0.074) (0.145)  Lagged HIV prevalence No Yes Yes Yes Yes  Trade openness No No Yes Yes Yes  Country effect No No No Yes Yes  Year effect No No No No Yes  Observations 828 828 828 828 828  Adj. R2 0.130 0.572 0.573 0.664 0.706 Note: Robust standard errors are reported in parentheses. Significance at the 10%, 5% and 1% level is indicated by *, ** and *** respectively. Table 2: OLS Estimates of the Relationship Between Income and HIV . (1) . (2) . (3) . (4) . (5) . Dependent variable: HIV incidence rate  Log (GDP p.c.) 0.424*** 0.120** 0.121** 0.011 0.212 (0.142) (0.055) (0.056) (0.074) (0.145)  Lagged HIV prevalence No Yes Yes Yes Yes  Trade openness No No Yes Yes Yes  Country effect No No No Yes Yes  Year effect No No No No Yes  Observations 828 828 828 828 828  Adj. R2 0.130 0.572 0.573 0.664 0.706 . (1) . (2) . (3) . (4) . (5) . Dependent variable: HIV incidence rate  Log (GDP p.c.) 0.424*** 0.120** 0.121** 0.011 0.212 (0.142) (0.055) (0.056) (0.074) (0.145)  Lagged HIV prevalence No Yes Yes Yes Yes  Trade openness No No Yes Yes Yes  Country effect No No No Yes Yes  Year effect No No No No Yes  Observations 828 828 828 828 828  Adj. R2 0.130 0.572 0.573 0.664 0.706 Note: Robust standard errors are reported in parentheses. Significance at the 10%, 5% and 1% level is indicated by *, ** and *** respectively. By controlling for country fixed effects, column (4) captures the within-country correlation between income and HIV. Controlling for country fixed effects leads to a quantitatively much smaller OLS coefficient on income that is statistically indistinguishable from zero. Column (5) adds to the regression time fixed effects. Including in the econometric model time fixed effects yields a larger OLS coefficient; however, standard errors also increase so that we cannot reject the null hypothesis that the OLS coefficient on income is equal to zero. OLS estimation of the impact that income has on HIV could be downward biased due to reverse causality. Negative reverse causality bias arises if HIV has a negative effect on labour productivity. Random measurement error in national accounts statistics implies that OLS estimates are attenuated towards zero. Table 3 reports the results of IV regressions that use the international commodity net-export price index as an instrument for income. We report the Kleibergen and Paap (KP) F-statistic and evaluate it against a critical value, adopted from Stock and Yogo (2005) that corresponds to the notion that 15% is the maximal rejection rate the researcher is willing to tolerate if the true rejection rate is 5%.9 A KP statistic exceeding this critical value implies that the maximal rejection rate is smaller than 15%, hence the actual size of the test is between the 5% and 15% levels. Table 3: IV Estimates of the Relationship Between Income and HIV . (1) . (2) . (3) . Pooled sample . Democracy . Autocracy . Second-stage dependent variable: HIV incidence rate  Log (GDP p.c.) 0.413 −0.209 0.746*** (0.454) (0.239) (0.250)  P-value: coefficient is same as in Col. II <0.01 First-stage dependent variable: Log (GDP p.c.)  Log(ComPI) 3.081*** 3.254*** 4.725*** (0.830) (0.918) (1.325)  First-stage Kleibergen–Paap F-Stat 9.6 12.55 12.71  Stock and Yogo critical value (15%) 8.96 8.96 8.96  Country + year fixed effects Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes  Trade openness Yes Yes Yes  First stage Adj. R2 0.858 0.977 0.835  Second Adj. R2 0.632 0.872 0.573 . (1) . (2) . (3) . Pooled sample . Democracy . Autocracy . Second-stage dependent variable: HIV incidence rate  Log (GDP p.c.) 0.413 −0.209 0.746*** (0.454) (0.239) (0.250)  P-value: coefficient is same as in Col. II <0.01 First-stage dependent variable: Log (GDP p.c.)  Log(ComPI) 3.081*** 3.254*** 4.725*** (0.830) (0.918) (1.325)  First-stage Kleibergen–Paap F-Stat 9.6 12.55 12.71  Stock and Yogo critical value (15%) 8.96 8.96 8.96  Country + year fixed effects Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes  Trade openness Yes Yes Yes  First stage Adj. R2 0.858 0.977 0.835  Second Adj. R2 0.632 0.872 0.573 Note: Robust standard errors are reported in parentheses. Significance at the 10%, 5% and 1% level is indicated by *, ** and *** respectively. Table 3: IV Estimates of the Relationship Between Income and HIV . (1) . (2) . (3) . Pooled sample . Democracy . Autocracy . Second-stage dependent variable: HIV incidence rate  Log (GDP p.c.) 0.413 −0.209 0.746*** (0.454) (0.239) (0.250)  P-value: coefficient is same as in Col. II <0.01 First-stage dependent variable: Log (GDP p.c.)  Log(ComPI) 3.081*** 3.254*** 4.725*** (0.830) (0.918) (1.325)  First-stage Kleibergen–Paap F-Stat 9.6 12.55 12.71  Stock and Yogo critical value (15%) 8.96 8.96 8.96  Country + year fixed effects Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes  Trade openness Yes Yes Yes  First stage Adj. R2 0.858 0.977 0.835  Second Adj. R2 0.632 0.872 0.573 . (1) . (2) . (3) . Pooled sample . Democracy . Autocracy . Second-stage dependent variable: HIV incidence rate  Log (GDP p.c.) 0.413 −0.209 0.746*** (0.454) (0.239) (0.250)  P-value: coefficient is same as in Col. II <0.01 First-stage dependent variable: Log (GDP p.c.)  Log(ComPI) 3.081*** 3.254*** 4.725*** (0.830) (0.918) (1.325)  First-stage Kleibergen–Paap F-Stat 9.6 12.55 12.71  Stock and Yogo critical value (15%) 8.96 8.96 8.96  Country + year fixed effects Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes  Trade openness Yes Yes Yes  First stage Adj. R2 0.858 0.977 0.835  Second Adj. R2 0.632 0.872 0.573 Note: Robust standard errors are reported in parentheses. Significance at the 10%, 5% and 1% level is indicated by *, ** and *** respectively. Column (1) of Table 3 reports IV estimates based on a sample that includes all SSA countries. The estimated coefficient on income is 0.41 and has a standard error of 0.45. Hence, the average effect of income on HIV in SSA countries is positive but not significantly different from zero. The KP F-statistic is above the Stock and Yogo critical value; and from the first stage we see that commodity price windfalls have a significant positive average effect on income. Columns (2) and (3) of Table 3 re-estimate the model for democracies and autocracies separately. For the first stage, the effects of commodity price windfalls on income are statistically significant in both democracies and autocracies. And the first stage KP F-statistics are higher than the Stock and Yogo critical value. In the second stage, we find that the estimated coefficient on income is negative in the sample of democracies. Quantitatively the coefficient on income is around −0.21; given the standard error, 0.24, we cannot reject the hypothesis that the coefficient is equal to zero. In SSA countries with autocratic regimes, the estimated coefficient on income is 0.75; its standard error is 0.25. We can reject the hypothesis that this coefficient is equal to zero at the 1% significance level. Further, we can reject the hypothesis that the coefficient on income in the autocracy sample is equal to the coefficient on income in the democracy sample at the 1% significance level. The coefficient on GDP per capita in column (3) of Table 3 implies that a 1% increase in income increases the HIV infection rate by around 0.7% in autocracies. Alternatively, this coefficient can be interpreted as a one standard deviation increase in GDP per capita leading to an about 0.6 standard deviation increase in the HIV infection rate. To interpret this result in terms of the actual number of people infected, we carry out some back-of-the-envelope calculations. For a small autocratic SSA country such as Equatorial Guinea, a 1% increase in GDP per capita leads to about 4,000 additional new infections of HIV. For larger autocracies such as Tanzania and Uganda, the approximate number of new infections of HIV that will result from a 1% increase in income is around 300,000 and 230,000, respectively.10 If we take a conservative perspective by using the two standard deviation lower bound of the point estimate of 0.75, the estimated number of deaths in Tanzania and Uganda from a 1% increase in income is around 96,000 and 74,000, respectively. 4.2 Robustness check to baseline results To see the effects of commodity price windfalls on HIV in democracies and autocracies, we present estimates of the reduced-form relationship. Columns (1) and (2) of Table 4 report least squares estimates of the within-country effect that commodity price windfalls have on HIV in democracies and autocracies, respectively. The estimates show that the effect of international commodity price windfalls on HIV infections is negative but not significantly different from zero in democracies; in autocracies the effect is positive and significantly different from zero at the 1% significance level. We can reject the hypothesis that the coefficient on the international commodity net-export price index in column (1) is equal to the coefficient in column (2) at the 1% significance level. Quantitatively, the coefficient in column (2) implies that a one standard deviation increase in the international commodity net-export price index is associated with an increase in the HIV incidence rate of about 0.5%. To interpret this result in terms of the actual number of people at risk, take Tanzania for example: a one standard deviation increase in the international commodity net-export price index is associated with about 200,000 new infections of HIV. Table 4: Reduced Regression and Alternative Income Data . (1) . (2) . (3) . (4) . Democracy . Autocracy . Democracy . Autocracy . Dependent variable: HIV incidence rate  Log (GDP p.c., WDI) −0.927 1.468**  Commodity windfall −0.680 3.479*** (1.171) (0.635) (0.833) (0.901)  P-value: coefficient <0.01 <0.01 First-stage dependent variable: Log (GDP p.c.)  Commodity windfall 0.731* 2.435** (0.405) (1.079)  First-stage Kleibergen–Paap F-Stat 3.26 5.295  Stock and Yogo critical value (15%) 8.96 8.96  Country + Year fixed effects Yes Yes  Lagged HIV prevalence Yes Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.992 0.932  Adj. R2 0.873 0.631 0.857 0.453 . (1) . (2) . (3) . (4) . Democracy . Autocracy . Democracy . Autocracy . Dependent variable: HIV incidence rate  Log (GDP p.c., WDI) −0.927 1.468**  Commodity windfall −0.680 3.479*** (1.171) (0.635) (0.833) (0.901)  P-value: coefficient <0.01 <0.01 First-stage dependent variable: Log (GDP p.c.)  Commodity windfall 0.731* 2.435** (0.405) (1.079)  First-stage Kleibergen–Paap F-Stat 3.26 5.295  Stock and Yogo critical value (15%) 8.96 8.96  Country + Year fixed effects Yes Yes  Lagged HIV prevalence Yes Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.992 0.932  Adj. R2 0.873 0.631 0.857 0.453 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Table 4: Reduced Regression and Alternative Income Data . (1) . (2) . (3) . (4) . Democracy . Autocracy . Democracy . Autocracy . Dependent variable: HIV incidence rate  Log (GDP p.c., WDI) −0.927 1.468**  Commodity windfall −0.680 3.479*** (1.171) (0.635) (0.833) (0.901)  P-value: coefficient <0.01 <0.01 First-stage dependent variable: Log (GDP p.c.)  Commodity windfall 0.731* 2.435** (0.405) (1.079)  First-stage Kleibergen–Paap F-Stat 3.26 5.295  Stock and Yogo critical value (15%) 8.96 8.96  Country + Year fixed effects Yes Yes  Lagged HIV prevalence Yes Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.992 0.932  Adj. R2 0.873 0.631 0.857 0.453 . (1) . (2) . (3) . (4) . Democracy . Autocracy . Democracy . Autocracy . Dependent variable: HIV incidence rate  Log (GDP p.c., WDI) −0.927 1.468**  Commodity windfall −0.680 3.479*** (1.171) (0.635) (0.833) (0.901)  P-value: coefficient <0.01 <0.01 First-stage dependent variable: Log (GDP p.c.)  Commodity windfall 0.731* 2.435** (0.405) (1.079)  First-stage Kleibergen–Paap F-Stat 3.26 5.295  Stock and Yogo critical value (15%) 8.96 8.96  Country + Year fixed effects Yes Yes  Lagged HIV prevalence Yes Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.992 0.932  Adj. R2 0.873 0.631 0.857 0.453 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Another robustness check that we have carried out is to use alternative national income data. Instead of using the real GDP per capita variable from PWT, we consider using data on real GDP per capita from WDI. Based on the WDI national income data, column (3) of Table 4 presents second-stage estimates for the sample of democracies; column (4) presents estimates for autocracies. We see that the second-stage coefficient on income is negative and not significantly different from zero in democracies; in autocracies it is positive and we can reject the hypothesis that the coefficient on income is equal to zero at the 5% significance level. Quantitatively, the size of the second-stage coefficients on income are larger (in absolute value) for regressions that are based on WDI data; however, the respective second stage coefficients are also associated with larger standard errors. The larger second stage standard errors are due to a more imprecise first stage fit between commodity price windfalls and income when data from the WDI are used. A further robustness check is to examine whether the results are sensitive to using an alternative democracy indicator. Columns (1) and (2) of Table 5 report estimates based on the democracy indicator of Przeworski et al. (2000). We see that the second-stage coefficient on income is negative and not significantly different from zero in democracies; in autocracies it is positive and we can reject the hypothesis that the coefficient on income is equal to zero at the 1% significance level. Quantitatively, the coefficients on income are similar to the benchmark estimates that are based on the polity democracy indicator. Table 5: Alternative Democracy and HIV Infection Data . (1) . (2) . (3) . (4) . Democracy . Autocracy . Democracy . Autocracy . Przeworski et al. (2000) . HIV incidence rate mortality based . Dependent variable: HIV incidence rate  Log (GDP p.c.) −0.541 0.669*** −47.313 0.719* (0.428) (0.253) (262.53) (0.517)  P-value: coefficient <0.01 <0.01 First-stage dependent variable: Log (real GDP per capita)  Commodity windfall 3.397*** 3.202*** 1.393* 3.468** (0.882) (1.108) (0.917) (1.562)  First-stage Kleibergen–Paap F-Stat 19.65 8.76 2.309 4.931  Stock and Yogo critical value (15%) 8.96 8.96 8.96 8.96  Country + Year fixed effects Yes Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.972 0.825 0.881 0.974  Adj. R2 0.828 0.528 0.808 0.686 . (1) . (2) . (3) . (4) . Democracy . Autocracy . Democracy . Autocracy . Przeworski et al. (2000) . HIV incidence rate mortality based . Dependent variable: HIV incidence rate  Log (GDP p.c.) −0.541 0.669*** −47.313 0.719* (0.428) (0.253) (262.53) (0.517)  P-value: coefficient <0.01 <0.01 First-stage dependent variable: Log (real GDP per capita)  Commodity windfall 3.397*** 3.202*** 1.393* 3.468** (0.882) (1.108) (0.917) (1.562)  First-stage Kleibergen–Paap F-Stat 19.65 8.76 2.309 4.931  Stock and Yogo critical value (15%) 8.96 8.96 8.96 8.96  Country + Year fixed effects Yes Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.972 0.825 0.881 0.974  Adj. R2 0.828 0.528 0.808 0.686 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Table 5: Alternative Democracy and HIV Infection Data . (1) . (2) . (3) . (4) . Democracy . Autocracy . Democracy . Autocracy . Przeworski et al. (2000) . HIV incidence rate mortality based . Dependent variable: HIV incidence rate  Log (GDP p.c.) −0.541 0.669*** −47.313 0.719* (0.428) (0.253) (262.53) (0.517)  P-value: coefficient <0.01 <0.01 First-stage dependent variable: Log (real GDP per capita)  Commodity windfall 3.397*** 3.202*** 1.393* 3.468** (0.882) (1.108) (0.917) (1.562)  First-stage Kleibergen–Paap F-Stat 19.65 8.76 2.309 4.931  Stock and Yogo critical value (15%) 8.96 8.96 8.96 8.96  Country + Year fixed effects Yes Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.972 0.825 0.881 0.974  Adj. R2 0.828 0.528 0.808 0.686 . (1) . (2) . (3) . (4) . Democracy . Autocracy . Democracy . Autocracy . Przeworski et al. (2000) . HIV incidence rate mortality based . Dependent variable: HIV incidence rate  Log (GDP p.c.) −0.541 0.669*** −47.313 0.719* (0.428) (0.253) (262.53) (0.517)  P-value: coefficient <0.01 <0.01 First-stage dependent variable: Log (real GDP per capita)  Commodity windfall 3.397*** 3.202*** 1.393* 3.468** (0.882) (1.108) (0.917) (1.562)  First-stage Kleibergen–Paap F-Stat 19.65 8.76 2.309 4.931  Stock and Yogo critical value (15%) 8.96 8.96 8.96 8.96  Country + Year fixed effects Yes Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.972 0.825 0.881 0.974  Adj. R2 0.828 0.528 0.808 0.686 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Columns (3) and (4) of Table 5 report results using Oster's mortality-based HIV infection data. We see that the second-stage coefficient on income is negative and not significantly different from zero in democracies; in autocracies it is positive and we can reject the hypothesis that the coefficient on income is equal to zero at the 10% significance level. 4.3 Agricultural versus mineral and fuel price shocks Intertemporal theories of consumption stipulate that consumption should respond more strongly to persistent income than to temporary variation income. If sex is a normal good, then persistent income shocks should have a larger effect on the demand for sex than transitory income shocks. As is well known, variations in agricultural commodity prices are more transitory in nature than those in mineral and fuel prices. Given this difference, it could be that our results regarding the HIV response in autocracies are mostly driven by those commodities where variations in prices are of persistent nature. In Table 6, we explore the effects of agricultural commodity price windfalls, and mineral and fuel commodity price windfalls, separately. Columns (1) and (2) show that both types of commodity price windfalls have a significant positive effect on income. Columns (3) and (4) show that whereas the coefficient on the agricultural commodity price index is positive and statistically indistinguishable from zero, the coefficient on the mineral and fuel price index is positive and significantly different from zero at the 1% level. Hence, our previous findings are driven mostly by variation in international commodity prices that are of persistent nature. Table 6: Agricultural Versus Mineral and Fuel Commodities on Income and HIV Dependent variable . (1) . (2) . (3) . (4) . Autocracy . Autocracy . Autocracy . Autocracy . Log (GDP p.c.) . Log (GDP p.c.) . HIV incidence rate . HIV incidence rate . Log(ComPI), Agricultural commodities 17.843*** 5.592 (6.657) (6.926) Log(ComPI), Mineral and fuel commodities 4.557*** 3.590*** (1.338) (0.936) Country + Year fixed effects Yes Yes Yes Yes Lagged HIV prevalence Yes Yes Yes Yes Trade openness Yes Yes Yes Yes Adj. R2 0.823 0.836 0.601 0.610 Dependent variable . (1) . (2) . (3) . (4) . Autocracy . Autocracy . Autocracy . Autocracy . Log (GDP p.c.) . Log (GDP p.c.) . HIV incidence rate . HIV incidence rate . Log(ComPI), Agricultural commodities 17.843*** 5.592 (6.657) (6.926) Log(ComPI), Mineral and fuel commodities 4.557*** 3.590*** (1.338) (0.936) Country + Year fixed effects Yes Yes Yes Yes Lagged HIV prevalence Yes Yes Yes Yes Trade openness Yes Yes Yes Yes Adj. R2 0.823 0.836 0.601 0.610 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Table 6: Agricultural Versus Mineral and Fuel Commodities on Income and HIV Dependent variable . (1) . (2) . (3) . (4) . Autocracy . Autocracy . Autocracy . Autocracy . Log (GDP p.c.) . Log (GDP p.c.) . HIV incidence rate . HIV incidence rate . Log(ComPI), Agricultural commodities 17.843*** 5.592 (6.657) (6.926) Log(ComPI), Mineral and fuel commodities 4.557*** 3.590*** (1.338) (0.936) Country + Year fixed effects Yes Yes Yes Yes Lagged HIV prevalence Yes Yes Yes Yes Trade openness Yes Yes Yes Yes Adj. R2 0.823 0.836 0.601 0.610 Dependent variable . (1) . (2) . (3) . (4) . Autocracy . Autocracy . Autocracy . Autocracy . Log (GDP p.c.) . Log (GDP p.c.) . HIV incidence rate . HIV incidence rate . Log(ComPI), Agricultural commodities 17.843*** 5.592 (6.657) (6.926) Log(ComPI), Mineral and fuel commodities 4.557*** 3.590*** (1.338) (0.936) Country + Year fixed effects Yes Yes Yes Yes Lagged HIV prevalence Yes Yes Yes Yes Trade openness Yes Yes Yes Yes Adj. R2 0.823 0.836 0.601 0.610 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. 5. Channels We now turn to the discussion of channels through which higher national income leads to an increase of the HIV infection rate in autocratic SSA countries. Our main hypotheses are that urbanisation is an important factor and the increase in urbanisation is supplemented in autocracies by lower public health expenditure shares in GDP. The first step in examining whether urbanisation and public health expenditures are channels through which income systematically affects HIV infections consists in regressing the latter variable on the former variables. The relevant results are presented in Table 7. Column (1) shows that the HIV infection rate is significantly positively correlated with urbanisation. This is in line with the stylised fact that, in autocracies, HIV infections occur at a higher rate in cities than in rural areas. We also find that public health expenditures are significantly negatively correlated with HIV infections, see column (2) of Table 7. This suggests that a decrease in the HIV infection rate goes hand in hand with an increase in public health expenditures. Table 7: Relation between HIV, Urbanisation and Public Health Expenditures in Autocracies. . (1) . (2) . (3) . Dependent variable is HIV infection rate  Urbanisation rate 7.660*** 7.690*** (2.178) (2.165)  Public health expenditure −3.607*** −3.336** (1.378) (1.579)  Country + year fixed effects Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes  Trade openness Yes Yes Yes  Adj. R2 0.632 0.602 0.602 . (1) . (2) . (3) . Dependent variable is HIV infection rate  Urbanisation rate 7.660*** 7.690*** (2.178) (2.165)  Public health expenditure −3.607*** −3.336** (1.378) (1.579)  Country + year fixed effects Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes  Trade openness Yes Yes Yes  Adj. R2 0.632 0.602 0.602 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Table 7: Relation between HIV, Urbanisation and Public Health Expenditures in Autocracies. . (1) . (2) . (3) . Dependent variable is HIV infection rate  Urbanisation rate 7.660*** 7.690*** (2.178) (2.165)  Public health expenditure −3.607*** −3.336** (1.378) (1.579)  Country + year fixed effects Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes  Trade openness Yes Yes Yes  Adj. R2 0.632 0.602 0.602 . (1) . (2) . (3) . Dependent variable is HIV infection rate  Urbanisation rate 7.660*** 7.690*** (2.178) (2.165)  Public health expenditure −3.607*** −3.336** (1.378) (1.579)  Country + year fixed effects Yes Yes Yes  Lagged HIV prevalence Yes Yes Yes  Trade openness Yes Yes Yes  Adj. R2 0.632 0.602 0.602 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Next, we show that urbanisation significantly increases with higher income in autocracies but not in democracies. For the autocracy sample, the coefficient on GDP per capita is positive and significantly different from zero at the 1% level, see column (1) of Table 8. On the other hand, the coefficient on GDP per capita is quantitatively small and statistically insignificant in the sample of democracies. We can also reject at the 1% significance level the hypothesis that the effects of income on urbanisation are the same in democracies and autocracies at the 1% significant level. With regard to public health expenditures, we find that national income has a significant negative effect on this variable in autocracies; in democracies the effect is insignificant. These results suggest that in autocracies increases in national income lead to an increase in urbanisation and lower public health expenditures. Table 8: Effect of Income on Urbanisation and Public Health Expenditures Second-stage dependent variable . Autocratic sample . Democratic sample . I . II . III . IV . Urbanisation rate . Public health expenditures . Urbanisation rate . Public health expenditures . Log (GDP p.c.) 0.075*** −0.151*** −0.002 0.038 (0.027) (0.034) (0.017) (0.037) First-stage dependent variable: Log (GDP per capita)  Log(ComPI) 4.708*** 4.379*** 3.126*** 3.254*** (1.349) (1.345) (0.974) (0.997)  First-stage Kleibergen–Paap F-Stat 12.653 11.059 12.068 12.598  Stock and Yogo critical value (15%) 8.96 8.96 8.96 8.96  Country + Year effects Yes Yes Yes Yes  Lagged HIV prevalence No Yes No Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.839 0.838 0.976 0.978  Second Adj. R2 0.966 0.772 0.995 0.570 Second-stage dependent variable . Autocratic sample . Democratic sample . I . II . III . IV . Urbanisation rate . Public health expenditures . Urbanisation rate . Public health expenditures . Log (GDP p.c.) 0.075*** −0.151*** −0.002 0.038 (0.027) (0.034) (0.017) (0.037) First-stage dependent variable: Log (GDP per capita)  Log(ComPI) 4.708*** 4.379*** 3.126*** 3.254*** (1.349) (1.345) (0.974) (0.997)  First-stage Kleibergen–Paap F-Stat 12.653 11.059 12.068 12.598  Stock and Yogo critical value (15%) 8.96 8.96 8.96 8.96  Country + Year effects Yes Yes Yes Yes  Lagged HIV prevalence No Yes No Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.839 0.838 0.976 0.978  Second Adj. R2 0.966 0.772 0.995 0.570 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Table 8: Effect of Income on Urbanisation and Public Health Expenditures Second-stage dependent variable . Autocratic sample . Democratic sample . I . II . III . IV . Urbanisation rate . Public health expenditures . Urbanisation rate . Public health expenditures . Log (GDP p.c.) 0.075*** −0.151*** −0.002 0.038 (0.027) (0.034) (0.017) (0.037) First-stage dependent variable: Log (GDP per capita)  Log(ComPI) 4.708*** 4.379*** 3.126*** 3.254*** (1.349) (1.345) (0.974) (0.997)  First-stage Kleibergen–Paap F-Stat 12.653 11.059 12.068 12.598  Stock and Yogo critical value (15%) 8.96 8.96 8.96 8.96  Country + Year effects Yes Yes Yes Yes  Lagged HIV prevalence No Yes No Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.839 0.838 0.976 0.978  Second Adj. R2 0.966 0.772 0.995 0.570 Second-stage dependent variable . Autocratic sample . Democratic sample . I . II . III . IV . Urbanisation rate . Public health expenditures . Urbanisation rate . Public health expenditures . Log (GDP p.c.) 0.075*** −0.151*** −0.002 0.038 (0.027) (0.034) (0.017) (0.037) First-stage dependent variable: Log (GDP per capita)  Log(ComPI) 4.708*** 4.379*** 3.126*** 3.254*** (1.349) (1.345) (0.974) (0.997)  First-stage Kleibergen–Paap F-Stat 12.653 11.059 12.068 12.598  Stock and Yogo critical value (15%) 8.96 8.96 8.96 8.96  Country + Year effects Yes Yes Yes Yes  Lagged HIV prevalence No Yes No Yes  Trade openness Yes Yes Yes Yes  First stage Adj. R2 0.839 0.838 0.976 0.978  Second Adj. R2 0.966 0.772 0.995 0.570 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. If urbanisation and public health expenditures are important channels through which income affects HIV infections in SSA countries, then one would expect to see the coefficient on income to become smaller (in absolute value) when including these variables in a multivariate regression model. Table 9 shows that, indeed, the conditional effect of income on HIV infections is quantitatively smaller. For example, column (1) of Table 9 shows that after controlling for urbanisation the coefficient on GDP per capita is 0.49. For comparison column (3) of Table 3 showed that, without controlling for urbanisation, the coefficient on GDP per capita is around 0.75. This suggests that up to one-third of the effect that income has on HIV is due to an increase in the urbanisation rate. Column (2) of Table 9 shows that when including in the multivariate regression model public health expenditures the coefficient on GDP per capita is around 0.045; given the standard error of 0.13 we cannot reject the hypothesis that this coefficient is equal to zero. Hence, when shutting down the effects that income has on urbanisation and public health expenditures, the (residual) effect of income on HIV infection is quantitatively small and statistically indistinguishable from zero. Table 9: Effect of Income on HIV in Autocracies Conditional on Urbanisation and Public Health Expenditures . (1) . (2) . Dependent variable is HIV infection rate  Log (real GDP per capita) 0.492*** 0.045 (0.174) (0.125)  Urbanisation rate 3.337*** 2.590** (1.085) (1.102)  Public health expenditure −3.320** (1.655) First-stage dependent variable: Log (real GDP per capita)  Commodity windfall 6.812*** 5.754*** (1.469) (1.227)  First-stage Kleibergen–Paap F-Stat 22.35 22.94  Stock and Yogo critical value (15%) 8.96 8.96  Country + Year fixed effects Yes Yes  Lagged HIV prevalence Yes Yes  Trade openness Yes Yes  First stage Adj. R2 0.859 0.878  Second Adj. R2 0.606 0.448 . (1) . (2) . Dependent variable is HIV infection rate  Log (real GDP per capita) 0.492*** 0.045 (0.174) (0.125)  Urbanisation rate 3.337*** 2.590** (1.085) (1.102)  Public health expenditure −3.320** (1.655) First-stage dependent variable: Log (real GDP per capita)  Commodity windfall 6.812*** 5.754*** (1.469) (1.227)  First-stage Kleibergen–Paap F-Stat 22.35 22.94  Stock and Yogo critical value (15%) 8.96 8.96  Country + Year fixed effects Yes Yes  Lagged HIV prevalence Yes Yes  Trade openness Yes Yes  First stage Adj. R2 0.859 0.878  Second Adj. R2 0.606 0.448 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. Table 9: Effect of Income on HIV in Autocracies Conditional on Urbanisation and Public Health Expenditures . (1) . (2) . Dependent variable is HIV infection rate  Log (real GDP per capita) 0.492*** 0.045 (0.174) (0.125)  Urbanisation rate 3.337*** 2.590** (1.085) (1.102)  Public health expenditure −3.320** (1.655) First-stage dependent variable: Log (real GDP per capita)  Commodity windfall 6.812*** 5.754*** (1.469) (1.227)  First-stage Kleibergen–Paap F-Stat 22.35 22.94  Stock and Yogo critical value (15%) 8.96 8.96  Country + Year fixed effects Yes Yes  Lagged HIV prevalence Yes Yes  Trade openness Yes Yes  First stage Adj. R2 0.859 0.878  Second Adj. R2 0.606 0.448 . (1) . (2) . Dependent variable is HIV infection rate  Log (real GDP per capita) 0.492*** 0.045 (0.174) (0.125)  Urbanisation rate 3.337*** 2.590** (1.085) (1.102)  Public health expenditure −3.320** (1.655) First-stage dependent variable: Log (real GDP per capita)  Commodity windfall 6.812*** 5.754*** (1.469) (1.227)  First-stage Kleibergen–Paap F-Stat 22.35 22.94  Stock and Yogo critical value (15%) 8.96 8.96  Country + Year fixed effects Yes Yes  Lagged HIV prevalence Yes Yes  Trade openness Yes Yes  First stage Adj. R2 0.859 0.878  Second Adj. R2 0.606 0.448 Note: Robust standard errors are reported in parentheses. Statistical significance at the 10%, 5% and 1% levels are indicated by *, ** and *** respectively. 6. Conclusion Both autocratic and democratic SSA countries experienced in the past decade significant increases in their national income. For example, during the period 1997 to 2007 the real GDP per capita of democratic SSA countries increased from about 1,700USD in 1997 to 2,700USD in 2007; in autocratic SSA countries GDP per capita increased from 1,300USD in 1997 to about 2,000USD in 2007. Both autocratic and democratic SSA countries thus experienced during 1997–2007 similar increases in their national income, at the rate of about 5% per annum. However, there was a remarkable difference in the evolutions of the urbanisation rate and HIV infections between autocracies and democracies: The increase in the urbanisation rate was about three times larger in autocracies than in democracies; HIV infections increased at a rate in autocracies that was about twice that in democracies.11 The above stylised facts suggest that the relationship between income and HIV infections may differ between autocratic and democratic SSA African countries; further, differences in the path of urbanisation stand out. The purpose of this paper was to examine in a rigorous way, based on panel data and instrumental variables estimation, the effects that national income has on HIV infections, and the role that urbanisation plays as a channel. Using variation in an international commodity net-export price index as an instrumental variable, we found that exogenous increases in national income lead to an increase in the HIV infection rate in autocratic SSA countries; in democracies the effect is insignificant. Differences in the response of urbanisation to exogenous income are one important factor driving this difference: whereas in autocratic SSA countries urbanisation significantly increased due to an exogenous income increase, no such effect was detected in democracies. The instrumental variables estimates showed that (increases in) urbanisation accounted for about one-third of the positive effect that national income had on HIV infections in autocratic SSA countries. Theoretically, the effects of urbanisation on HIV infections are ambiguous: in cities the costs of finding multiple sexual partners are lower but so are the (unit) costs of preventing the spread of the disease; and migrants in cities, being away from spouses and family monitor, are also more likely to be involved in transactional sex. However, in principle governments can take action to prevent the spread of the disease and in cities such actions will be more cost effective. Whether governments choose to implement such policies is an open question. The empirical evidence provided in this paper showed that in autocratic SSA countries exogenous increases in national income lead to a significant decline in public health expenditures. This suggests that autocratic SSA regimes neglected the public health care system and thus failed to prevent the spread of HIV. Supplementary material Supplementary material is available at Journal of African Economies online. Footnotes 1 " Source: ‘The Global HIV/AIDS Epidemic’, Kaiser Family Foundation, 5 August 2014. Available at: http://kff.org/global-health-policy/fact-sheet/the-global-hivaids-epidemic/. 2 " In general, overcrowding in urban areas can facilitate the transmission of diseases and infections (Fay et al., 2005; Cutler et al., 2006). However, urbanisation may help to reduce HIV transmission if there is better access to medical facilities in urbanised areas. We will show later in the paper that urbanisation and HIV infections are positively correlated in autocratic SSA countries. 3 " Whether governments choose to implement such policies, i.e., policies that are beneficial to a broad spectrum of society, depends on the incentives that governments face. Political institutions (among other factors) shape these incentives (North, 1981). 4 " See http://www.unctad.org/Templates/Page.asp?intItemID=1584/&lang=1. 5 " Please see Appendix Table B for the summary statistics of the polity2 score for each SSA country. 6 " Detailed data on time to death are difficult to generate, particularly in developing countries, since it requires knowing (roughly) the time of infection. For this reason, she uses the available data from developed countries and it is shown that the trend is actually similar with Africa. See Oster (2012, p. 1035) for details. 7 " Also see Table C in the Appendix for the summary of the definitions and sources for each variable. 8 " For example, high prevalence of tropical disease could reduce income and have adverse effects on health (McArthur and Sachs, 2001). Gallup et al. (1999) show that in the tropics, income is generally lower and human health is adversely affected by tropical climate. 9 " This follows from the suggestion of Stock et al. (2002). 10 " This is computed using the population size in 2007. For instance, Tanzania's population in 2007 was around 40 million. Using the semi-elasticity of 0.75, we compute the approximate number of additional infections as 40 million × 0.0075 = 300,000. 11 " Specifically, the urbanisation rate (HIV infection rate) increased in democratic SSA countries from 0.33 (0.62) in 1997 to 0.35 (0.79) in 2007; in autocratic SSA countries the figures are 0.32 (0.92) and 0.38 (1.44), respectively. References Acemoglu D. , Robinson J. A. ( 2012 ) ‘Why Nations Fail: The Origins of Power’, in Prosperity, and Poverty , 1st edn. New York : Crown Business . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Acemoglu D. , Robinson J., Verdier T. ( 2004 ) ‘ Kleptocracy and Divide-and-rule: A Model of Personal Rule ’, Journal of the European Economic Association , 2 : 162 – 92 . Google Scholar Crossref Search ADS WorldCat Ahlburg D. , Jensen E. ( 1998 ) ‘The economics of the commercial sex industry’, in Ainsworth M., Fransen L., Over M. (eds), Confronting AIDS: Evidence from the Developing World, European Commission/World Bank . Luxembourg : Office for Publications of the European Communities , pp. 147 – 74 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Arezki R. , Brueckner M. ( 2012 a) ‘ Commodity Windfalls, Polarization, and Net Foreign Assets ’, Journal of International Economics , 86 : 318 – 26 . Google Scholar Crossref Search ADS WorldCat Arezki R. , Brueckner M. ( 2012 b) ‘ Commodity Windfalls, Democracy, and External Debt ’, Economic Journal , 122 : 848 – 66 . Google Scholar Crossref Search ADS WorldCat Avelino G. , Brown D., Hunter W. ( 2005 ) ‘ The Effects of Capital Mobility, Trade Openness, and Democracy on Social Spending in Latin America, 1980–1999 ’, American Journal of Political Science , 49 : 625 – 41 . Google Scholar Crossref Search ADS WorldCat Baird S. , Garfein R., McIntosh C., Özler B. ( 2012 ) ‘ Effect of a Cash Transfer Programme for Schooling on Prevalence of HIV and Herpes Simplex Type 2 in Malawi: A Cluster Randomised Trial ’, The Lancet , 379 : 1320 – 9 . Google Scholar Crossref Search ADS WorldCat Besley T. , Kudamatsu M. ( 2008 ) ‘Making autocracy work’, in Helpman E. (ed.), Institutions and Economic Performance . Cambridge : Harvard University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Biggs B. , King L., Basu S., Stuckler D. ( 2010 ) ‘ Is Wealthier Always Healthier? The Impact of National Income Level, Inequality, and Poverty on Public Health in Latin America ’, Social Science & Medicine , 71 : 266 – 73 . Google Scholar Crossref Search ADS WorldCat Bolnik J. , Kayuni H., Mabala R., McGranahan G., Mitlin D., Nkhoma S., Oucho J., Sabri A., Sabry S., Satterthwaite D., Swiling M., Tacoli C., Tambulasi R., van Donk M. 2006 . ‘A Pro Poor Urban Agenda For Africa: Clarifying Ecological and Development Issues for Poor and Vulnerable Populations', Human Settlements Discussion Paper 2. Brown D. , Hunter W. ( 2004 ) ‘ Democracy and Human Capital Formation ’, Comparative Political Studies , 37 : 842 – 64 . Google Scholar Crossref Search ADS WorldCat Brueckner J. ( 1990 ) ‘ Analyzing Third World Urbanization: A Model with Empirical Evidence ’, Economic Development and Cultural Change , 3 : 587 – 610 . Google Scholar Crossref Search ADS WorldCat Brueckner M. , Ciccone A. ( 2010 ) ‘ International Commodity Prices, Growth and the Outbreak of Civil War in Sub-Saharan Africa ’, Economic Journal , 120 : 519 – 34 . Google Scholar Crossref Search ADS WorldCat Brueckner M. ( 2012 ) ‘ Economic Growth, Size of the Agricultural Sector, and Urbanization in Africa ’, Journal of Urban Economics , 71 : 26 – 36 . Google Scholar Crossref Search ADS WorldCat Brueckner M. , Ciccone A., Tesei A. ( 2012 ) ‘ Oil Price Shocks, Income, and Democracy ’, Review of Economics and Statistics , 94 : 389 – 99 . Google Scholar Crossref Search ADS WorldCat Burke M. , Gong E., Jones K. ( 2015 ) ‘ Income Shocks and HIV in Africa ’, The Economic Journal , 125 : 1157 – 89 . Google Scholar Crossref Search ADS WorldCat Case A. , Deaton A. ( 2005 ) ‘ Health and Wealth Among the Poor: India and South Africa Compared ’, American Economic Review , 95 : 229 – 33 . Google Scholar Crossref Search ADS WorldCat Clark S. ( 2004 ) ‘ Early Marriage and HIV Risks in sub-Saharan Africa ’, Studies in Family Planning , 35 : 149 – 60 . Google Scholar Crossref Search ADS PubMed WorldCat Cutler D. , Deaton A., Lleras-Muney A. ( 2006 ) ‘ The Determinants of Mortality ’, Journal of Economic Perspectives , 20 : 97 – 120 . Google Scholar Crossref Search ADS WorldCat Davis J. , Henderson J. V. ( 2003 ) ‘ Evidence on the Political Economy of the Urbanization Process ’, Journal of Urban Economics , 53 : 98 – 125 . Google Scholar Crossref Search ADS WorldCat de Walque D. , Dow W. H., Nathan R., Abdul R., Abilahi F., Gong E., Isdahl Z., Jamison J., Jullu B., Krishnan S., Majura A., Miguel E., Moncada J., Mtenga S., Mwanyangala M. A., Packel L., Schachter J., Shirima K., Medlin C. A. ( 2012 ) ‘ Incentivising Safe Sex: A Randomised Trial of Conditional Cash Transfers for HIV and Sexually Transmitted Infection Prevention in Rural Tanzania ’, BMJ Open , 2 : e000747 . Google Scholar Crossref Search ADS PubMed WorldCat Deaton A. ( 1999 ) ‘ Commodity Prices and Growth in Africa ’, Journal of Economic Perspectives , 13 : 23 – 40 . Google Scholar Crossref Search ADS WorldCat Dinkelman T. , Lam D., Leibbrandt M. ( 2008 ) ‘ LinkinG Poverty and Income Shocks to Risky Sexual Behaviour: Evidence From a Panel Study of Young Adults in Cape Town ’, South African Journal of Economics , 76 : S52 – 74 . Google Scholar Crossref Search ADS PubMed WorldCat Fay M. , Leipziger D., Wodon Q., Yepes T. ( 2005 ) ‘ Achieving Child-health-related Millennium Development Goals: The Role of Infrastructure ’, World Development , 33 : 1267 – 84 . Google Scholar Crossref Search ADS WorldCat Gallup J. L. , Sachs J. D., Mellinger A. ( 1999 ) ‘ Geography and Economic Growth ’, International Regional Science Review , 22 : 179 – 232 . Google Scholar Crossref Search ADS WorldCat Glaeser E. L. ( 2014 ) ‘ A World of Cities: The Causes and Consequences of Urbanization in Poorer Countries ’, Journal of the European Economic Association , 12 : 1154 – 99 . Google Scholar Crossref Search ADS WorldCat Heston A. , Summers R., Aten B. 2009 . Penn World Table Version 6.3. Center for International Comparisons of Production, Income and Prices, University of Pennsylvania. Henderson V. ( 2003 ) ‘ The Urbanization Process and Economic Growth: The So-what Question ’, Journal of Economic Growth , 8 : 47 – 71 . Google Scholar Crossref Search ADS WorldCat Kaufman R. R. , Segura-Ubiergo A. ( 2001 ) ‘ Globalization, Domestic Politics, and Social Spending in Latin America ’, World Politics , 53 : 553 – 87 . Google Scholar Crossref Search ADS WorldCat Kohler H. P. , Thornton R. L. ( 2012 ) ‘ Conditional Cash Transfers and HIV/AIDS Prevention: Unconditionally Promising? ’, The World Bank Economic Review , 26 : 165 – 90 . Google Scholar Crossref Search ADS PubMed WorldCat Kudamatsu M. ( 2012 ) ‘ Has Democratization Reduced Infant Mortality in Sub-Saharan africa? Evidence From Micro Data ’, Journal of the European Economic Association , 10 ( 6 ): 1294 – 1317 . Google Scholar Crossref Search ADS WorldCat Levine D. , Rothman D. ( 2006 ) ‘ Does Trade Affect Child Health ’, Journal of Health Economics , 25 : 538 – 54 . Google Scholar Crossref Search ADS PubMed WorldCat Lin F. Q. , Sim N. C. ( 2013 ) ‘ Trade, Income and Baltic Dry Index ’, European Economic Review , 59 : 1 – 18 . Google Scholar Crossref Search ADS WorldCat Lin F. Q. , Sim N. C. ( 2015 ) ‘ Exports, HIV Incidence and the Baltic Dry Index: Further Evidence from Sub-Saharan Africa ’, Economics Letters , 126 : 35 – 39 . Google Scholar Crossref Search ADS WorldCat LoPiccalo K. , Robinson J., Yeh E. 2012 . ‘Income, Income Shocks, and Transactional Sex', The Oxford Handbook of the Economics of Prostitution. Lurie M. N. , Williams B. G., Zuma K., Mkaya-Mwamburi D., Garnett G. P., Sweat M. D., Gittelsohn J., Karim S. S. ( 2003 ) ‘ Who Infects Whom? HIV-1 Concordance and Discordance among Migrant and Non-migrant Couples in South Africa ’, Aids , 17 : 2245 – 52 . Google Scholar Crossref Search ADS PubMed WorldCat Marshall M. , Jaggers K. 2009 . Polity IV Project: Dataset Users’ Manual. Center for Global Policy, George Mason University. McArthur J. W. , Sachs J. D. 2001 . Institutions and Geography: Comment on Acemoglu, Johnson, and Robinson (2000), NBER Working Paper, 8114. McGuire J. W. ( 2006 ) ‘ Democracy, Basic Service Utilization, and Under-5 Mortality: A Cross-national Study of Developing States ’, World Development , 34 : 405 – 25 . Google Scholar Crossref Search ADS WorldCat Oster E. ( 2005 ) ‘ Sexually Transmitted Infections, Sexual Behavior and the HIV/AIDS Epidemic ’, Quarterly Journal of Economics , 120 : 467 – 515 . OpenURL Placeholder Text WorldCat Oster E. ( 2012 ) ‘ Routes of Infection: Exports and HIV Incidence in Sub-Saharan Africa ’, Journal of the European Economic Association , 10 : 1025 – 58 . Google Scholar Crossref Search ADS WorldCat Padró i Miquel G. ( 2007 ) ‘ The Control of Politicians in divided Societies: The Politics of Fear ’, The Review of Economic Studies , 74 ( 4 ): 1259 – 74 . Google Scholar Crossref Search ADS WorldCat Potts D. ( 2009 ) ‘ The Slowing of sub-Saharan Africa's Urbanization: Evidence and Implications for Urban Livelihoods ’, Environment and Urbanization , 21 ( 1 ): 253 – 9 . Google Scholar Crossref Search ADS WorldCat Pritchett L. , Summers L. H. ( 1996 ) ‘ Wealthier is Healthier ’, The Journal of Human Resources , 31 : 841 – 68 . Google Scholar Crossref Search ADS WorldCat Przeworski A. , Alvarez M., Cheibub J., Limongi F. ( 2000 ) Democracy and Development: Political Institutions and the Well-Being of the World, 1950–1990 , 1st edn. Cambridge : Cambridge University Press . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Rajan K. , Kennedy J., King L. ( 2013 ) ‘ Is Wealthier Always Healthier in Poor Countries? The Health Implications of Income, Inequality, Poverty, and Literacy in India ’, Social Science & Medicine , 88 : 98 – 107 . Google Scholar Crossref Search ADS WorldCat Robinson J. , Yeh E. ( 2011 ) ‘ Transactional Sex as a Response to Risk in Western Kenya, American Economic ’, Journal: Applied Economics , 3 : 35 – 64 . Google Scholar Crossref Search ADS WorldCat Ruhm C. J. ( 2007 ) ‘ A Healthy Economy Can Break Your Heart ’, Demography , 44 : 829 – 48 . Google Scholar Crossref Search ADS PubMed WorldCat Skoufias E. ( 2003 ) ‘ Economic Crises and Natural Disasters: Coping Strategies and Policy Implications ’, World Development , 31 : 1087 – 1102 . Google Scholar Crossref Search ADS WorldCat Stasavage D. ( 2005 ) ‘ Democracy and Education Spending in Africa ’, American Journal of Political Science , 49 : 343 – 58 . Google Scholar Crossref Search ADS WorldCat Stock J. H. , Wright J. H., Yogo M. ( 2002 ) ‘ A Survey of Weak Instruments and Weak Identication in Generalized Method of Moments ’, Journal of Business and Economic Statistics , 20 : 518 – 29 . Google Scholar Crossref Search ADS WorldCat Stock J. H. , Yogo M. ( 2005 ) ‘Testing for weak instruments in linear IV regression’, in Andrews D. W. K., Stock J. H. (eds), Identication and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg . Cambridge : Cambridge University Press . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Swidler A. , Watkins S. C. ( 2007 ) ‘ Ties of Dependence: AIDS and Transactional Sex in Rural Malawi ’, Studies in family planning , 38 : 147 – 62 . Google Scholar Crossref Search ADS PubMed WorldCat Tavares J. , Wacziarg R. ( 2001 ) ‘ How Democracy Affects Growth ’, European Economic Review , 45 : 1341 – 78 . Google Scholar Crossref Search ADS WorldCat Todaro M. P. 1997 . Urbanization, Unemployment, and Migration in Africa: Theory and Policy. Policy Research Division Working Paper 104. New York: Population Council. Tornell A. , Lane P. R. ( 1999 ) ‘ The Voracity Effect ’, American Economic Review , 89 ( 1 ): 22 – 46 . Google Scholar Crossref Search ADS WorldCat World Development Indicators , 2014 . Online Database. www.worldbank.org Author notes † " We want to thank Professor Markus Brueckner for his help and support to the paper. Faqin Lin acknowledges that the work is supported by National Natural Science Foundation of China (71773148 & 71503281). The work is also supported by Program for Innovation Research in Central University of Finance and Economics and Young Elite Teacher Project of Central University of Finance and Economics. © The Author 2017. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com.
TI - Income from International Commodity Price Windfalls and HIV Infections in sub-Saharan Africa
JO - Journal of African Economies
DO - 10.1093/jae/ejx020
DA - 2017-11-01
UR - https://www.deepdyve.com/lp/oxford-university-press/income-from-international-commodity-price-windfalls-and-hiv-infections-g29CnU0Ljs
SP - 607
VL - 26
IS - 5
DP - DeepDyve
ER -