Democracy, institutions and famines in developing and emerging countries

Abstract

Abstract The paper has a two-fold objective. First, it theoretically and empirically analyses the effects of democracy on famine mortality. Second, it examines the role played by other institutional/governance factors. The econometric exercises realised with data on a group of emerging and developing countries confirm the validity of Amartya Sen's ‘democracy prevents famine’ argument. Moreover, two main institutional indicators, computed by the World Bank, ‘control of corruption’ and ‘government effectiveness’, are found to be negatively correlated with famine mortality, suggesting that the policy environment, the level of bureaucracy and governmental capacity to take prompt decisions are relevant for reducing famine mortality. These factors are important also among countries with the same political regime.

Re´sume´ Ce travail a un double objectif. D'abord, il analyse théoriquement et empiriquement les effets de la démocratie sur la mortalité causée par la famine. Deuxièmement, il examine le rôle joué par d'autres facteurs institutionnels et de gouvernance. Les études économétriques, basées sur des données d'un groupe de pays émergents, confirment la validité de l'argument que « la démocratie empêche la famine » d'Amartya Sen. En outre, deux principaux indicateurs institutionnels, « la lutte contre la corruption » et « l'efficacité du gouvernement », mesurés par la Banque mondiale, se trouvent être négativement liés à la mortalité par famine, ce qui suggère que l'environnement politique, le niveau de bureaucratie, et la capacité du gouvernement à prendre des décisions rapides, sont des facteurs déterminants pour la réduction de la mortalité par famine. Ces facteurs sont également importants parmi les pays ayant le même régime politique.

Keywords:

• Famine
• democracy
• institutions
• developing countries
• cross-country analysis

1. Introduction

The paper engages in the debate on political institutions and famine in low- and middle-income countries. Famines are catastrophic phenomena which cause large social and economic damages to a country. Thus, the presence of famines is a symptom of a lower development level as well as a constraint to future development.

In order to tackle famine, formal and informal institutions can play a major role. Institutions are here viewed as rules and social norms that shape and affect agents' behaviour, so structuring social interaction (Knight 1992). This article firstly concentrates on the effect of democracy on the likelihood of a country experiencing famine and, broadly speaking, its effects on famine mortality. We will review the debate concerning the ‘democracy prevents famines’ thesis elaborated by the economist Amartya Sen. According to this argument, genuine democracies have specific features which prevent them from experiencing a famine even during periods of economic crisis. However, these mechanisms, according to Sen, are not in place to counter chronic problems such as poverty, hunger and inequality (Sen 1983, 1999, Easaw and Savoia 2009, Burchi and Tridico 2010). The validity of this thesis will be tested through an econometric analysis carried out with data from a large sample of emerging and developing countries.

Then we will extend our study to other factors strictly related to the quality of institutions and the governance. In particular we will examine the effects of factors relating to governance, as they are reflected in indicators elaborated by the World Bank – ‘voice and accountability’, ‘control of corruption’, ‘political stability’ and ‘government effectiveness’ – on famine mortality. Another assumption that we intend to test is whether the possible effect of these institutions-related indicators replaces the effect of democracy or, on the contrary, integrates with it. Our hypothesis is that democracy is crucial for famine prevention, but also that some autocratic regimes, equipped with other adequate institutions (for example, a functioning bureaucracy, low corruption or a stable political system), can succeed in preventing famines. While there has been an increasing number of investigations into the causal relationship between institutions and economic growth (North and Thomas 1981, Knack and Keefer 1995), to our best knowledge the relationship between political institutions and famine mortality has never been examined by using quantitative methods with a large panel of developing and middle-income countries.

The remainder of the article is organised as follows: in the next two sections we discuss the concept of famine and review the debate on democracy, political institutions and famine; in the following section we present the conceptual framework, the data and the results of a set of econometric models we propose; and in the final section we draw our conclusions.

2. What is a famine?

The concept of famine is a very complex and multi-faceted one. As argued by de Waal (2000), it is impossible to define it properly because it is a multidisciplinary concept. Therefore, we might define it according to the main outcomes. In this sense, a famine is a phenomenon involving ‘acute starvation and a sharp increase of mortality’, distinct from chronic hunger, which involves ‘sustained nutritional deprivation on a persistent basis’ (Dreze and Sen, 1989, p. 7). In a more comprehensive way, de Waal (2000) identifies four major outcomes of a famine: (1) hunger; (2) impoverishment; (3) social breakdown; and (4) mortality.

The definition of famine is not independent of the approach followed in order to study it. For a long time the debate on hunger and famine has been dominated by the thought of Thomas Malthus (1798), who focused attention on two key variables: food availability and population. Given the fact that the growth rate of population is much higher than that of food production, he argued that the future lack of enough food per capita would cause famines and starvation and concluded that famine be viewed as a sharp decline in food availability in a country or region.

It was only at the beginning of 1980s that Amartya Sen's entitlement approach contributed to challenging this perspective, drawing attention to access to food. ‘The entitlement approach concentrates on each person's entitlements to commodity bundles including food and views starvation as resulting from a failure to be entitled to any bundle with enough food’ (Sen 1981, p. 434). Entitlements depend mainly on two elements: (1) personal endowments, which are the resources held by a person such as house, work and land; (2) the set of commodities accessible to the person through trade and production, that is, the ‘exchange entitlement mapping’ (Sen 1981, p. 435). Famine can be the consequence of a decline of endowments or a decline in the exchange entitlement mapping; for instance, a sharp reduction in the price of the commodity produced reduces that person's capacity to buy food.

The validity of the above theoretical framework was tested in three major famines: Bengal (1943); the Wollo Province of Ethiopia (1973); and Bangladesh (1974) (Sen 1981). The main finding is that in none of the three countries was there a decline in available food during the year of the crisis.1 This shows that famines cannot be fully understood by simply focusing on per capita food availability.

The entitlement approach has been very influential in addressing new famines. It provided empirical evidence that famines can occur even without any decline in food availability, that ultimately we should look at the available means that people or households have to access food and, finally, that we need to analyse the conditions of different social groups. Quoting Sen (1981, p. 434, emphasis original), ‘starvation is a matter of some people not having enough food to eat and not a matter of there being not enough food to eat’.

De Waal (1989a, 1989b) contributes to a further understanding of famines by pointing out that starvation rarely leads directly to death. More often, starvation occurs together with the spread of diseases like diarrhoea and measles; these two combined factors are then responsible for excess mortality in famines. The author uses the Darfur famine of 1984–1985 as a case study for this theory, discovering that ‘food and income, with the sole exception of access to milk, completely failed to predict mortality’ (de Waal 1989b, p. 21). As a conclusion, he argues that ‘famine mortality in Darfur in 1985 was truly disease-driven and not starvation-driven’ (de Waal 1989b, p. 24). This leads to a broader definition of famine, as provided by de Waal (2000) and highlighted earlier in this section. Moreover, Dreze and Sen (1989), too, have further expanded Sen's theoretical framework for studies of famine by extending the focus from ‘food entitlements’ to overall entitlements,2 thus recognising that other factors such as access to medical care and drinking water are also important explanatory factors of famine-related mortality.

Finally, while famine and food insecurity are two rather different phenomena, there exists a close relationship between them. All other factors held constant, countries or regions where hunger and poverty are widespread are also more vulnerable to external shocks. In these areas, manmade and natural disasters or a sudden increase in food prices are more likely to lead to famine. On the other hand, in countries experiencing frequent famines the number of people falling into the hunger trap tends to increase. For example, Shaw (2007, ch. 41) pointed out that the repeated emergencies (some of which have been labelled ‘famines’) that have hit Ethiopia between 1986 and 2003 have caused a dramatic increase in the proportion of people chronically food insecure and, in particular, child malnutrition. Similarly, Young et al. (2005) and Young and Maxwell (2009) stress the effect of conflicts and drought on livelihood stability. As these authors argue, in the case of Sudan, famines have caused depletion of assets and natural resources and a general impoverishment, which, in turn, had the deleterious effect of raising long-run problems of food insecurity and nutritional distress. As Young et al. (2005, p. 26) emphasise, ‘famine and food insecurity are an integral feature of livelihoods in Darfur, shaping the way people manage their resources, livelihood goals and livelihood strategies’.

3. The role of democracy and political institutions

In the past there was a fatalistic attitude towards famines because they were believed to be caused by man-made or natural disasters such as droughts and floods, which in turn reduced food production. The entitlement approach, by shifting attention to socio-economic factors and breaking down the analysis by groups, has contributed to the identification of new determinants of famines, recognising a much wider role of the state and other agents. As highlighted very effectively by Devereux (2000, p. 27), ‘famines occur because they are not prevented: they are allowed to happen’. The central focus of the present sub-section is on ‘who’ (which institutions? which organisations?) should intervene and ‘why’ (what are the incentives?); the ‘how’ question (which policies?) will be discussed but only briefly.

The second feature of a democracy is a free and independent media. While the extent of freedom of expression might depend on many other factors, which vary from country to country, a broader independence of the media should be an intrinsic element of a genuine democracy. When a disaster occurs, a free media plays a crucial informative role because it provides information on the symptoms of the crisis both to individuals and to political parties. Furthermore, news media can put direct pressure on governments and other actors in charge of taking action. Dynamic journalism, independent from the political sphere, is able to raise awareness of the situation by acting as an early warning system, to openly criticise the work of the government and, finally, to demand an immediate intervention.

The third intrinsic element of a well-functioning democracy is the presence of a vibrant civil society. This includes people gathering and mobilising in order to protest against the current conditions and exercise their voice. In addition, in a genuine democracy, non-governmental organisations (NGOs), grass-roots organisations and other voluntary associations play a number of roles: (1) an informative one because they often work with the poor, who are those usually hit by famine; (2) they can supply goods and services to reduce the effects of the crisis; (3) as active organisations, they should be able to put pressure on political institutions as well as being intermediaries between the poor and the formal institutions. While a dynamic civil society is not exclusive to democracies, we believe that it can freely exercise all its roles only in this political environment.

Several scholars have stressed the potential role of democracy in famine prevention (for example, Bardhan 1999, Banik 2007, Osmani 2007). Using the words of Osmani (2007, p. 10), ‘two attributes of democracy are at work here – viz., the scope for open debate as an accountability-demanding mechanism and the presence of election as an accountability-enforcing mechanism’. Similarly, Bardhan (1999, p. 102) argues:

Democracy helps development through the accountability mechanisms it installs for limiting the abuse of executive power and provides a system of periodic punishments for undesirable government interventions in the economy and rewards for desirable interventions. … Accountability mechanisms are particularly important in averting disasters.

As a consequence, famines are caused by a widespread entitlement failure, but the latter is mainly the result of an ‘institutional failure’.

Both cross-country evidence and evidence based on looking at any one single country before and after a change in political system have been shown to support the idea that democracy prevents famines. Comparisons between Zimbabwe and Botswana, on the one hand, and Sudan and Ethiopia, on the other, in the period 1979–1984, fall into the first category of evidence. During this period, many disasters caused a decline in food production by 38 per cent and 17 per cent respectively in Zimbabwe and Botswana and by 11 per cent and 12 per cent in Sudan and Ethiopia. Though the scale of natural disasters was much higher in the first group of countries, famine occurred only in the latter group because of the lack of political incentives for action (Sen 1999).

China and India, traditionally put in the same cluster because of their size, population density, geographical location and recent transition to market economy, are compared along the same lines (Sen 1999). While China at the beginning of the 1960s was still ineffective in preventing famines and during 1958–1961 actually experienced the most dramatic famine in the contemporary world, India had managed to avoid the occurrence of famines since the 1947 transition to democracy. The authoritarian regime in China did not allow for an adequate flow of information around the country and did not make the government accountable towards its citizens. With reference to Ethiopia and China, D'Souza (1994, p. 373) argues that ‘the absence of democracy, the lack of independent media, the prevalence of draconian censorship resulting in a culture of fear – all these played a pivotal and disastrous role in the famines described here.’

Finally, the inter-temporal comparison of India provides additional support for the protective role of democracy. Both before and after 1947, the country experienced a significant reduction in food availability. However, after the Great Bengal Famine of 1943, no major famine was registered. Investigative journalism and an active opposition party always imposed early intervention on the government so that natural disasters did not move inevitably into famine.

Ram (1995) offers one of the few in-depth analyses with respect to the historical influence of the media in India. Since the end of the nineteenth century there was already an independent press, which provided relevant information during emergencies such as the 1918 food crisis. In this situation, it delivered timely and numerous reports, which generated protests around the country. However, the press was not listened to by the national government, which could and did hide the gravity of the crisis and suppress all protests. This happened because the independent press was marginal and the nationalist press protected the government and because of the lack of the reinforcing mechanisms of a democracy. A free and independent press was crucial in famine prevention after 1947. This conclusion is reinforced by the results of the quantitative analyses carried out by Besley and Burgess on a panel of Indian states between 1958 and 1992. The authors find that in the face of crises the government was more prone to protect its citizens through a set of relief policies in states where the media was more highly developed (Besley and Burgess 2002).

Finally, with reference to the dramatic droughts hitting the Indian state of Maharashtra in 1960–1963, Dreze (1995) identifies three factors that played a substantial role in avoiding a disaster: the Famine Codes, established at the end of the nineteenth century; the ‘public pressure from political parties, the media, [and] voluntary agencies’ (Dreze 1995, p. 155); and a well-planned intervention.3 The Famine Codes offered important guidelines for the implementation of adequate policies, but without the mechanisms ensured by a democratic regime the government would probably have ignored them, as often happened in the pre-democracy period.

As a conclusion to this large reflection, Sen argued that ‘there has never been a famine in a functioning multiparty democracy’ (Sen 1999, p. 178).

3.1. Critiques and counter-critiques

The idea that democracy plays an essential role in preventing famines has received some critiques, which challenge both previous empirical evidence and the causal mechanisms that might underlie the relationship between democracy and famine.

Among others, Brass (1986), de Waal (2000), Myhrvold-Hanssen (2003) and Rubin (2009b) argue that there was indeed a famine in the Indian state of Bihar in 1966–1967. Myhrvold-Hanssen (2003) also points out that following Sen's definition of famine as a phenomenon involving excess starvation and mortality and a sudden decline in average calorie intake with large variations across areas and groups, the crisis that occurred in Bihar might well be labelled a famine. This would undermine the validity of the statement that no democracy has experienced a famine as well as the identification of India as a fully successful case of famine prevention after the transition to a democratic regime. With respect to the case of Bihar, Myhrvold-Hanssen (2003) stresses that the Indian media, although free and independent, did not provide reliable information, while other commentators such as Brass (1986) identify the democratic system as one possible cause of the famine. The first effects of the crisis were visible just before the elections and the ruling party had a political disincentive to recognise the famine. What drives the actions of different agents, according to Brass, is self interest and so there could be egoistic reasons in democratic contexts that also could lead a government not to declare and therefore not to intervene ahead of a famine.

There are also other controversial cases of possible famines in seemingly democratic countries. Among these, de Waal (1989a, 2000) focuses on Bangladesh in 1974, Sudan in 1986–1988 and Ireland in 1945–1949. However, as also stressed by the author, these cases are exceptional because Bangladesh ‘was democratic and liberal in name only’ and South Sudan and Ireland were not considered as parts of Sudan and the United Kingdom respectively. Finally, more recently, Rubin (2008b, 2009a, 2009b) has argued that Malawi and Niger experienced famines in 2002 and 2005 respectively, though they were defined as democratic countries by several commentators. Following the line of thought of Devereux and Tiba (2007), Rubin (2008b) indicates the fragility of the state and its dependence on international assistance as the main political causes of the famine in Malawi. The denial of the existence of the famine and the general inactivity of the government were, instead, responsible for the famine in Niger in 2005 (Rubin 2009a).

The above critiques focus on possible exceptions to Sen's argument according to which ‘there has never been a famine in a functioning multiparty democracy’ (Sen 1999, p. 178). Many of these controversies are around the definition of famine and democracy and the estimates of mortality rates and calorie intakes. While these debates are still open, we believe that there are cases in which Sen's statement was violated. However, the empirical evidence presented in this sub-section does not undermine the validity of the broad statement that democracy prevents famine argument.4 It helps in identifying possible exceptions in which a fairly democratic country did not succeed in preventing famines.

The points raised by these authors also help in moving away from a narrow definition of democracy as ‘electoral democracy’ and in going beyond a simple democracy/autocracy dichotomy. In addition to an analysis of ‘democratic institutions’, there is a need to examine how these institutions are ‘used’ (defined as ‘democratic practices’ by Dreze and Sen 2003) within a country. While the former are institutional arrangements that constitute the environment in which ‘democratic ideals’5 can be pushed forward, ‘democratic practices’ depend ‘on the extent of political participation, the awareness of the public … and popular organisations’ (Dreze and Sen 2002, p. 6). This way, we can move towards an idea of democracy as ‘government by discussion’ (Buchanan 1954, Sen 2003).

Another group of commentators argues that famines are political processes, which have both victims and beneficiaries (Edkins 1996, Keen 1994). Some (national or international) interest groups may obtain advantages from the occurrence of a famine, and this is true in autocratic as well as democratic states. Some democratic states in the Horn of Africa have often used violence (for example, through legislation prohibiting protests) to protect the interests and the assets of the political elites, leaving the poor and vulnerable segments of the society dying of starvation and diseases (Edkins 1996, Keen 1994). According to this school of thought, violence is not adequately taken into account in Sen's theoretical framework and it is necessary to examine the details of each step of the political process leading to a specific famine rather than constructing a grand theory like the democracy prevents famine theory of famine causation (Edkins 2007, p. 57).

Finally, the critiques addressed by Brass (1986), Rubin (2009b) and Plümper and Neumayer (2009) challenge the fundamentals of Sen's thought: democratic countries might even have lower incentives to respond to crises than authoritarian ones. The reason lies in the possible ‘political blame’, which is the practice of public institutions of passing the responsibility to other ones and because of the low political weight of the poor, who are the main victims of famines. According to these authors, these are the main causes of government inactivity, especially in the Bihar and Niger famines described above.

With a small amount of empirical evidence and theoretical arguments that seem to help the understanding in a few cases rather than building an alternative model of behaviour of democratic governments facing a crisis, we feel that these critiques do not undermine the validity of Sen's arguments. However, they offer interesting insights into the idea of concentrating more on the institutional arrangements within and outside democracies. The example of Malawi in 2002 focuses attention on state capacity to implement adequate policies within a short time. Whether or not a country is democratic influences the system of incentives likely to make the government accountable. However, it is then necessary to analyse the effectiveness and efficiency of these institutions both in general terms and for the specific purpose of preventing famines. These further elements are incorporated in the following quantitative analysis.

4. An empirical exercise

4.1. Model specification

In this section we present an empirical analysis, which has two main objectives: (1) to verify with cross-country regressions based on recent data whether or not democratic countries are more likely to limit famine mortality; and (2) to analyse to what extent the quality of political institutions matters within a group of countries homogenous for political regime. The model proposed here is similar to that developed by Plümper and Neumayer (2009), but refers only to emerging and low-income countries.

The main model is expressed by the following equation:

Famine mortality is a function of the degree of democracy (DEMO), the proportion of people affected (affected), the amount of calories per capita (cal), the amount of drinkable water per capita (water), the intensity of civil conflicts (intconf), the (logarithm of) population (logPop) and population density (popdens). This way, we analyse the effect of democracy on the number of deaths due to famines, controlling for these relevant factors. We acknowledge the possible bias due to the fact that both democracy and famine may be endogenous to other factors such as per capita GDP. However, preliminary analysis of correlation (for example, no systematic linear relationship between democracy and GDP) as well as post-estimation tests and sensitivity analyses allow us to conclude that this bias should not be substantial.

The second group of models is expressed by the general equation:

Equation (2) replaces the variable DEMO with other institutional variables. Voice and accountability (voice), control of corruption (ControlCorr), political stability (stability) and government effectiveness (goveff) will be separately included in the models.6 It would be important to add another crucial variable related to people's (especially the ‘hungry’) participation in decision making because, as argued in the previous section, a good climate created by a well-functioning democracy does not lead automatically to a wider participation of people in community and political life. In particular, we recognise that the populations targeted by relief and development interventions should take ownership of actions to respond to famines, rather than just being passive recipients of food aid or cash transfers. However, given the complex nature of these important factors, reliable data is missing.

Then we will investigate the effect of the institutional indicators on famine mortality within countries with the same political regimes, that is, separately for democratic and autocratic countries. In this way, it is possible to verify, for example, whether institutional factors explain lower famine mortality, given the same form of government. A system where corruption is not widely diffused within the society and does not take a systemic form might favour the functioning of institutions, whether or not they are labelled democratic (Dreze and Sen 2002). Within democracies, a highly corrupted system in which illegal or immoral behaviours and norms are routinely followed might increase the likelihood of famine by undermining democratic practices, that is, people's real opportunity to be informed, to express their opinions, to vote freely and to participate in public life. Similarly, countries characterised by overwhelming bureaucracy or by political instability might have lower performances in reducing famine mortality.

4.2. Data

Data were originally collected for 110 low-income and emerging countries for the period 1972–2006. Due to missing values in some key variables as well as problems in managing data concerning countries that have been created, merged, or divided during this period, the final number of observations for the estimation of Equation (1) was brought down to 2905, including 102 countries for, on average, 28.5 years.

The dependent variable is a count variable: number of deaths during a famine. It is very difficult to measure famine-related mortality; thus any source of this type of information has surely some degree of error. The problem of a lack of commonly shared definition of famine exacerbates the problem. According to us, the best source of these data is the EM-DAT database developed by the Catholic University of Leuven. This database is used by international organisations such as the World Bank, the World Health Organisation and the United Nations International Strategy for Disaster Reduction Secretariat, and was utilised in similar quantitative studies by Neumayer and Plümper (2007), Rubin (2008a) and Plümper and Neumayer (2009). Moreover, an in-depth investigation carried out by Adger et al. (2004) concluded that the EM-DAT data, especially those concerning famine mortality, are sufficiently reliable. An alternative source would come from the paper written by Devereux (2000), where the author reviews different studies on famine and compiles estimates of famine mortality in each of them. The main problem is that this work covers only the famines with more than 1,000 deaths and famines that occurred until 1998.7 As a result, in the present paper we employ EM-DAT data. The only exception concerns the famine which occurred in Bangladesh in 1974, where data are not available in the EM-DAT database but were taken from Devereux (2000).

Data concerning the number of people affected – defined as injured, homeless and all ‘people requiring immediate assistance during a period of emergency’8 – were taken from the same source.9 The final variable – the proportion of people affected by the famine – was obtained by dividing this number by the size of the population (UN Secretariat 2007).

Democracy is measured by the reverse of the political rights index computed by the Freedom House. It ranges from one (less rights/democracy) to seven (more rights/democracy). As suggested by the Freedom House, countries with a political rights index above or equal to five are defined as democratic, while those with an index below five are autocratic.

The other main variables are four governance indicators computed by the World Bank, which indicate the quality of institutions (Kaufmann et al. 2007). The first is ‘voice and accountability’, which takes into account various aspects related to the political process, civil liberties and the independence of the media. The second is the ‘control of corruption’ index, which reflects the reverse of the level of corruption within national institutions as perceived by government officials and the private sector. The ‘political stability and absence of violence’ index, instead, combines ‘several indicators measuring perceptions of the likelihood that the government in power will be destabilised or overthrown by unconstitutional or violent means’ (Kaufmann et al. 2004, p. 254). Finally, ‘government effectiveness’ describes the ability of governments to effectively deliver public services and make policies. All these indicators are obtained by aggregating a series of variables and by finally transforming them in standardised indices varying approximately between –2.5 and +2.5. Data are available only for the following years: 1996; 1998; 2000; 2002–2006. This substantially reduces the number of observations used to estimate the effects of institutions on famine mortality. The correlation analysis of the institutional variables (not reported) shows that ‘voice and accountability’ is highly correlated to democracy,10 while ‘control of corruption’ is the institutional indicator with the weakest correlation to democracy.

The source of data for calories and water per capita is the FAO, the Faostat and Aquastat databases respectively. In particular, the first variable is central to the interpretation of famines as food (or calories) availability decline criticised here. Instead, the intensity of conflicts is taken as a measure of the severity of civil conflicts and varies between zero (0–24 deaths) and two (1,000 deaths or more). Finally, population data are taken from the UN Secretariat (2007) and divided by the total area of the countries concerned (FAO/FAOSTAT n.d.) to calculate population density.

4.3. The estimates

The estimates were obtained through negative binomial regression because: (1) the dependent variable – famine mortality – is a count variable; and (2) there is overdispersion.11

The estimates in Model (1) of Table 1 show that the democracy index has a highly significant (negative) effect on the number of deaths during famines. This result is coherent with the theoretical framework and with the previous findings of Plümper and Neumayer. In contrast, Rubin (2008a), with different model specifications and alternative measures of famine, found no systemic correlation between democracy and occurrence of famines. Among the other explanatory variables, the proportion of people affected and per capita water availability are the only significant determinants.

Table 1. Negative binomial estimates of famine mortality: full sample.

	(1) with DEMO	(2) with Voice	(3) with ControlCorr	(4) with Stability	(5) with Goveff
Constant	−4.807*	−4.726	−10.450**	−5.106	−12.626***
Democracy	−0.770***	–	–	–	–
Voice	–	−0.802	–	–	–
ControlCorr	–	–	−3.793***	–	–
Stability	–	–	–	−1.004	–
Goveff	–	–	–	–	−3.239***
Affected	24.977***	71.443	67.633	67.752	89.752
Calorie	0.001	−0.002	−0.002	−0.003*	−0.001
Water	−0.000***	−0.000	−0.000	−0.000	−0.000**
Intconflict	0.297	−0.206	−0.455	−1.003	−0.153
Logpop	0.263	0.470	0.602*	0.556*	0.714**
Popdens	0.000	−0.006**	−0.005*	−0.005**	−0.008**
1/lnalpha	6.144***	5.047	4.912	5.050	5.005
Alpha	465.949***	155.630***	135.947***	155.953***	149.227***
Log pseudo-likelihood	−415.466	−74.107	−72.878	−74.109	−73.474
Wald chi2	274.31***	34.01***	65.23***	33.60***	74.58***
N	2905	462	460	470	475
Note: *** = significant at the 0.01 level; ** = significant at the 0.05 level; * = significant at the 0.1 level. The estimates were carried out with robust standard errors; clustered standard errors led to the same levels of significance for all the variables.

We then studied whether or not other institutions-related factors influence famine mortality. The estimates are based only on 460 to 475 observations, clearly reducing the reliability of results; however, we think they can still be used to sort out the general relationship between two or more phenomena. The estimates in Models (2) to (5) outline that only control of corruption and government effectiveness have significant (and negative) coefficients. We can conclude from these results that effective and efficient governments together with effective rules and transparent institutions may reduce the likelihood of facing a famine, whether or not a country is ‘democratic’ in a narrow sense.

To understand better the previous findings, we ran separate regressions for democratic and autocratic countries. The results – presented in Table 2 – suggest that in the sample of democratic countries only control of corruption and government effectiveness are significant negative determinants of famine mortality. This indicates again that among countries generally defined as democratic there are differences in the way they manage to keep corruption low and to adopt effective policies, and this is finally reflected in lower famine mortality among other things. Though the group of democratic states registers much higher mean values in the two institutional indicators, such indicators have a great variability across countries in this group (Table 3).

Table 2. Negative binomial estimates of famine mortality: selected sub-samples.

	Democratic countries				Autocratic countries
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Constant	−1.934	−7.211	−5.102	−19.101***	−6.753	−5.812	−6.523	−11.910***
Voice	−2.087	–	–	–	−3.333**	–	–	–
ControlCorr	–	−7.517***	–	–	–	−5.619***	–	–
Stability	–	–	0.876	–	–	–	−4.416***	–
Goveff	–	–	–	−8.843***	–	–	–	−3.151**
Affected	46.154	38.393	23.433	44.258	123.250	81.204	49.204**	142.207
Calorie	−0.004**	−0.005**	−0.004***	−0.005***	−0.003	−0.001	−0.003***	−0.002
Water	−0.000***	−0.004***	−0.000**	−0.000***	−0.000	−0.000	−0.000***	−0.000
Intconflict	−0.830	−0.018	−0.852	−1.126**	1.017	0.468	−2.873*	0.698
Logpop	0.770*	0.942**	0.950**	1.824***	0.558	0.229	0.598	0.731
Popdens	−0.005**	−0.006**	−0.004**	−0.009***	−0.025**	−0.028***	−0.027***	−0.022***
1/lnalpha	3.787	3.476	3.788	3.480	5.749	5.643	5.716	5.757
Alpha	44.104	32.346	44.159	32.464	313.724	282.324	303.584	316.358
Log pseudo-likelihood	−36.133	−34.497	−36.329	−34.281	−33.938	−33.629	−33.855	−33.959
Wald chi2	72.79***	57.01***	84.17***	39.20***	56.77***	55.95***	55.52***	53.36***
N	190	197	200	202	272	263	270	273
Note: *** = significant at the 0.01 level; ** = significant at the 0.05 level; * = significant at the 0.1 level. The estimates were carried out with robust standard errors; clustered standard errors (not reported) led to the same levels of significance for all the variables.

Table 3. Descriptive statistics of institutional indicators, by political regime.

Democratic countries
Variable	Obs	Mean	Std dev.	Min.	Max.
Voice	369	0.190	0.408	−1.160	0.990
ControlCorr	390	−0.251	0.567	−1.350	1.480
Stability	384	−0.090	0.698	−1.993	1.401
Goveff	396	−0.194	0.557	−2.520	1.470
Autocratic countries
Variable	Obs	Mean	Std dev.	Min	Max
Voice	476	−1.004	0.530	−2.300	0.200
ControlCorr	468	−0.733	0.571	−2.090	1.620
Stability	477	−0.890	0.869	−3.073	1.017
Goveff	478	−0.788	0.590	−2.270	1.010

All the institutional variables are significant determinants of famine mortality in autocratic countries – Models (5) to (8) in Table 2. This suggests that non-democratic states can still have a wide set of institutional arrangements that can prevent famines or reduce famine mortality. Voice and accountability and political stability are significant predictors within autocratic states; also, their mean value is much larger in democratic states but has a large variability in those which are non-democratic (Table 3).

The latter finding does not contradict Sen's argument on the protective role of democracy. Putting his argument in other words, democracy is likely to be sufficient for famine prevention, but it is not strictly necessary. De Waal (2000) gives four examples of successful attempts of non-democratic governments to prevent famines: (1) Communist China in the period immediately after the 1940s; (2) Sudan, Ethiopia and Somalia during the 1970s; (3) Kenya in 1984; and (4) Tigray in northern Ethiopia during the mid-1980s. The point is that in the above cases, ‘famine measures are a privilege rather than a right’ (de Waal 2000, p. 18). Again quoting de Waal:

There can be anti-famine commitments and anti-famine programmes in the absence of democratic accountability, but an anti-famine contract requires the interested party – the people – to have some capacity to enforce the bargain. Liberal political systems provide a number of mechanisms that can help people to do that. In authoritarian systems, the only recourse is protest, either armed or unarmed. (de Waal 2000, p. 18)

The lack of re-enforcing mechanisms can seriously undermine the sustainability of these results. For example, in the case of China outlined above, in 1959–1962 there was the biggest famine that had ever occurred in the contemporary era; also in the case of most of the African authoritarian regimes, the system of anti-famine measures collapsed immediately after the 1970s.

Among the other results, it is worth noting that there is no systemic significant negative relationship between calories per capita and famine mortality, as one would expect if following the food availability decline approach to famine. Regarding the other variables, the coefficients of water per capita and, even more so, population density are often significant and of the expected sign.

4.4. Discussion

This study provides empirical evidence in support of the ‘democracy prevents famine’ hypothesis. It should also be stressed that the results were obtained on a sample of low-income and emerging countries: given that high-income countries have rarely experienced famines since 1972, the final estimates are likely to underestimate the overall effect of democracy.

The second finding is that a more in-depth analysis should go beyond the system of political incentives existing in different political regimes and move into the analysis of the quality of institutions. The policy environment, the level of bureaucracy, governmental capacity to take decisions and implement them in a short period and other factors linked to governance are, in a second stage, crucial for entitlement protection and promotion. Both the variables ‘government effectiveness’ and ‘control of corruption’ are significant (negative) determinants of the number of deaths due to famine.

Finally, enlightened authoritarian governments with adequate political institutions rather than democratic ones can still avoid the occurrence of famines even in the face of human catastrophes. However, validity of these achievements in the long run might be at constant risk due to the lack of re-enforcing mechanisms ensured by people's right to express their voice, a multi-party system, free and independent media and regular elections.

5. Concluding remarks

The main aim of the paper was to review the existing literature on the relationship between democracy and famine and to empirically test the assumption that the former plays an important role in preventing the latter. Moreover, the paper intended to investigate whether the quality of institutions matters for famine relief, and finally whether its potential role operates as a substitute for, or interacts with, democracy.

The findings of the empirical analyses conducted on a panel of low- and middle-income countries largely support the ‘democracy prevents famines’ hypothesis elaborated by Amartya Sen during the 1980s. The variable expressing the degree of democracy has a largely significant and negative effect on famine mortality.

The following exercises, carried out on separate samples of democratic and autocratic countries, show that only ‘control of corruption’ and ‘government effectiveness’ of the four governance indicators are significant in the first sample, while they are all significant in the second. This suggests that enlightened authoritarian governments with adequate, though not democratic, political institutions can still avoid the occurrence of famines even in the face of a dramatic crisis caused by a natural or manmade disaster. However, the sustainability of these achievements might be undermined by the lack of re-enforcing mechanisms ensured by the individual's right to express opinions, a free and independent media and periodic elections. In one sentence, ‘famine measures are a privilege rather than a right’ in authoritarian regimes (de Waal 2000, p. 18).

Finally, these results contribute to add new insights and evidence to sort out the effect of democracy and political institutions on famine prevention. Further empirical research is needed in the future to understand in greater depth the causal mechanisms through which these relationships operate. In particular, it would be useful to include additional determinants of famine mortality, both related (for example, people's political participation) and unrelated to institutions and governance, to verify whether the core results change.

Notes

It should be highlighted that some authors have contested these empirical findings. Goswami (1990), Ó Grada (2008) and Majumdar (2010) argued that a decline in food availability did occur in the Bengal famine, while Basu (1984) and Kumar (1990) arrived at the same conclusion in the case of the Wollo and Bangladesh famines respectively. For a comprehensive review, see Dyson (1996, ch. 3).

For a detailed discussion of food versus overall entitlements, see Burchi and De Muro (2010).

This was based on restoring the purchasing power of the main victims through public employment. Similar policies were successfully implemented in other Indian crises, Botswana, Cape Verde and Zimbabwe (Dreze and Sen 1989, ch. 7).

Most of these authors explicitly recognise that their critiques do not touch the core of Sen's view (for example, de Waal [2000]).

Democratic ideals ‘include political characteristics that can be seen to be intrinsically important in terms of the objective of democratic social living, such as freedom of expression, participation of the people in deciding on the factors governing their lives, public accountability of leaders and an equitable distribution of power’ (Dreze and Sen 2002, p. 6).

Also in this case, we are aware that the quality of institutions may be itself endogenous to the model (Besley 2006). However, there are no useful techniques to adequately handle the problem.

For those famines for which we could compare EM-DAT data with those highlighted by Devereux (2000), we see no systematic underestimation or overestimation of the mortality figures in the first source.

Source: EM-DAT glossary, available from, http://www.emdat.be/glossary/9.

For a discussion of the limits of EM-DAT data, especially concerning the estimates of the number of people affected by a drought, see Below et al. (2007).

The Pearson's rho is above 0.90 for all years. That is why ‘voice and accountability’ can be used as a proxy for democracy (see Section 4.1).

The standard deviation of the dependent variable is more than 37 times larger than the mean: this suggests the use of negative binomial regression rather than Poisson regression.