1. Introduction
Extensive research has been conducted on the topic of how media circulation affects political accountability and government policy. Theory predicts that for a higher share of media receivers, political accountability and hence government expenditures increase. Besley & Burgess (2002) give additional insight into this topic by analyzing the impact of media circulation on government responsiveness to falls in food production and crop flood damage in Indian states. The authors use the extent of public food distribution and calamity relief as proxies for government responsiveness. In addition to media factors, political and economic factors are introduced as potential determinants of policies. The predictions of the theory are underlined by the results of the paper: Government responsiveness increases with a higher amount of media users within a state. Further, political factors are also relevant determinants, whereas, economic factors are of low importance.
In the following, the paper will be critically assessed within these sections. First, the theory, the propositions as well as the empirical strategy are introduced and compared to discussions in class. Second, the results of the paper are outlined. Third, the empirical strategy as well as the results will be analyzed and compared to prior research and theory. Finally, a short conclusion and outlook will be given.
2. Theory, Propositions and Empirical Strategy
The theoretical two-period model of Besley & Burgess (2002) is based on several assumptions. There are vulnerable and non-vulnerable citizens. Part of the vulnerable citizens are needy, meaning that they suffer after a shock in a certain period and that public action would improve their situation. Moreover, there are altruistic, selfish, and opportunistic incumbents. For the purpose of the study, opportunistic incumbents are most interesting because they put in effort if reelection chances increase thereby, whereas altruistic incumbents always put in the maximal effort level and selfish incumbents always put in zero effort. Finally, citizens base their voting decisions on different criteria. Only vulnerable citizens are influenced in their decision by the effort put in by incumbents. Vulnerable citizens who are not needy receive information about an incumbents’ effort from the media, needy citizens do in addition directly experience the effort level. It is assumed that uninformed vulnerable citizens do not vote. Non-vulnerable citizens decide who to vote for solely by means of ideological reasons.
The model assumes that an opportunistic incumbent elected in the first period chooses the effort level in order to maximize reelection chances for the subsequent period. In the second period however, only altruistic incumbents will put in effort to help the needy because there is no incentive anymore for opportunistic incumbents to put in effort since a third period and therefore also reelection chances are nonexistent. As vulnerable citizens do not vote for selfish incumbents, they only vote for an incumbent in case they learned about effort put in by this incumbent in the prior period.
The above outlined model gives the following predictions: An opportunistic incumbent increases his effort level if voters have better access to media because in this case the population learns about the effort. Moreover incumbents show more effort if turnout is higher in elections, if a larger fraction of the population is vulnerable, and if the incumbent’s advantage over the challenger is low. The third section below will show to what extent these propositions are supported by the results of the two authors.
The above outlined model is largely in line with the theory discussed in class. The voting model is based on the assumptions that politicians maximize their own utility and voters can punish politicians by not reelecting them. The model presented in class has several characteristics in common with the model by Besley & Burgess (2002). Both models only consist of two periods, politicians chose effort in first period in order to maximize reelection chances in the second period and put in no effort in the second period. However, there are also some differences in the models. The model presented in class does not know altruistic politicians who put in effort even though this does not maximize their utility. Moreover the model presented in class does not distinguish between vulnerable and non-vulnerable citizens which are influenced by different factors in their voting decision. Additionally, in the model presented in class uninformed citizens do vote which is not the case in the model by Besley & Burgess (2002). Furthermore, the model by Besley & Burgess focuses only on incumbents’ effort while the other model also takes incumbents’ competences into account. Finally, factors such as individual sympathy towards a candidate and popularity of a candidate only influence voting decisions in the model presented in class and are neglected by the two authors.
Besley & Burgess (2002) test their model using data from 16 Indian states in the period 1958-1992. The effort put in by the incumbents is measured in public food distribution and calamity relief expenditure. Food grain production and flood damage are used as indicators for the need for government intervention. The effect of media development on government effort is examined by observing data on newspaper circulation. Newspaper circulation statistics are considered on an aggregate level as well as disaggregated in English, Hindi and other language newspapers. It is assumed that newspapers in languages other than English and Hindi are mainly read by local, vulnerable citizens. Additionally, the study takes electoral turnout and political competitiveness as proxies for political factors as well as state income, urbanization, population and revenue from the center as proxies for economic factors into account. Hence, Beasley & Burgess do not only use media determinants of government activism but further introduce economic and political factors to improve the model.
3. Results
Based on the above described model and the application of the empirical data, the paper presents the following results. First, food production shortfall and food damage are highly correlated with extreme weather conditions such as intense rainfall or drought. This relationship supports the use of the two variables as basis for the measure of need for government actions as the two factors are subject to exogenous components. Second, stemming on the regression analysis, which puts economic factors and government activism into relation, the results suggest that these economic variables exert a limited influence on government activism and responsiveness.
Third, newspaper circulation and government responses show a positive correlation. The resulting effect of newspaper circulation on the two government responses introduced in the model are significantly positive as 1% increase in circulation can lead to 2.4% gain in public food distribution and 5.5% rise in calamity relief expenditures. Therefore, states with advanced media systems offer more protection for vulnerable citizens because at a given level of food damage or production shortfall, states with higher newspaper circulation will exert a higher level of government responsiveness.
Fourth, in terms of electoral turnout and political competition the analysis produces mixed results for government activism. Turnouts in previous elections do not influence the government responsiveness. Further, whereas increased political competition effects higher public food distribution, it does not affect calamity relief expenditures in a similar way. Moreover, public food distribution also varies with timing of the election. An explanation for this result may be the nature of the policy measure: Public food distribution is often used as a highly visible and politicised mean of dealing with food shortages (Radhakrishna & Subbarao, 1997). Electoral turnout and political competition exhibit positive correlations towards government responses, as both, higher electoral turnout and political competition lead to higher responsiveness for a given fall of food production or food damage. This result is consistent with the idea that increased election threat and competition pressure lead to higher responsiveness. Nevertheless, also in this case the political effects are more significant for food distribution than for calamity relief.
Fifth, the Indian language landscape consists of a variety of different (local) languages and Hindi as well as English being the languages used in national coverage. The analysis of language effects in the paper yields no effect on government responses national wide for English and Hindi, whereas newspaper circulation in other languages manifests positive effect on government responsiveness. This result is reasonable considering the access to language education of the vulnerable population. To summarize, most of the theories developed within the model can be verified with the results from the empirical studies. Whereas economic development seems to be unimportant, election with trends of political competition set some incentive for government responses. On top of that, media development can assist in the stimulation of government responsiveness.
4. Evaluation of Empirical Strategy
When evaluating results of an economic study, it always is crucial to check whether the strategy used is appropriate to infer causality and correlation. Especially reverse causality and omitted variables can undermine those. In their analysis, Besley & Burgess (2002) use multiple regression in order to show that increased media circulation comes along with higher government responsiveness following falls in food production or crop flood damage. In a multiple regression, it is always possible that unobservable factors influence outcomes that can't be controlled for. In this context, however, the two authors go beyond prior research by not only including media but also economic and political factors as possible influences on the government expenditures. This allows them for example to exclude the possibility that increased government action is fostered by state income and increased media is just another side-effect that comes along with a higher state income. In this example, the variable state income would just be affecting both, expenditures and media circulation and hence, impact of media circulation would be refuted. To summarize, whenever variables that affect cause as well as outcome are excluded, the results have to be analysed with care. By including a broad base of relevant variables, the authors lower the possibility for occurrence of this omitted variable bias and hence improve the empirical strategy.
Another specific example in this context is that the authors instrumented newspaper circulation with newspaper ownership. Here again the concern is that high circulation states differ from low circulation states in other variables such as political interest. Accounting for the instrumentation further improves the results. In essence, the authors have considered the most relevant variables in their research. Nevertheless, several omitted variables such as the influence of non-profit organisations or foreign influence could have further improved the results of the study. Significant influence of non-profit organisations or foreign countries could, for example impact both, media circulation as well as government expenditures and hence weaken the correlation between these two factors.
Next to omitted variables, reverse causality is a key concept to control for. In this context, reverse causality would mean that, in fact, government support would lead to increased newspaper circulation in those states and not the other way round as predicted by the results. Unfortunately, Besley & Burgess (2002) do not control for this factor. It could very well be possible that politicians encourage newspaper circulation in regions where they have put in a lot of effort in order to increase their chances for re-election. To sum up, omitted variables bias and reverse causality are always of key importance in an economic model and when accounted for can significantly strengthen the results.
As analysed above, the empirical strategy used by Besley & Burgess (2002) is convincing. The final question to be discussed is whether the results are consistent with theoretical predictions and how the results relate to prior research. First, the results of the study do indeed match the predictions of theory on the effects of media circulation on government policy: The two authors show that government expenditures do increase with a higher media circulation. Second, the results are further supported by prior research. Strömberg (2004) uses a voting model to analyse the effect of radio introduction on relief funds and concludes that counties with a large share of radio users receive more government support. Reinikka & Svensson (2005) use evidence from a newspaper campaign in Uganda in order to show that schools that are easier accessible with newspapers receive more government funds. Snyder & Strömberg (2010) estimate the impact of news coverage on actions by the political representatives in the respective region. Their key finding further supports the predictions of the theory as areas with a higher coverage of the politician receive more federal support. Although, Besley & Burgess do not focus on coverage of media but rather on the circulation, the results of Snyder & Strömberg give additional insights into the interplay of media and policy. In essence, the results of Besley & Burgess follow the theory and are supported by previous research.
5. Conclusion
Having used a strong empirical model, the authors show that a higher media circulation leads to an increase in government responsiveness. In addition, political factors are an important determinant of support. These results do not only underline theory but are also supported by broad prior research on the topic of media and policy. The impact of these results is widespread as it is implied that voters that are lacking access to newspapers are disadvantaged with respect to government support. Obviously, this decreases social welfare. In the combination with other research, the above outlined papers can be of crucial importance in fighting the discrimination against regions lacking access to newspapers. In addition, further research could shed additional light on this topic. For example, it could be analysed whether the theory also holds true in countries other than India. Further, it would be interesting to see whether newspaper circulation also has an impact on other kinds of government support such as education. Overall, the two authors present strong results with respect to the effects of media access on policy and thereby not only impact future policy making but also lay a strong foundation for further research on this topic.

6. References

Besley, T., & Burgess, R.S.L. (2002), The Political Economy of Government Responsiveness: Theory and Evidence from India. Quarterly Journal of Economics, 117 (4), 1415-51.
Prat, A., & Strömberg, D. (2013). The Political Economy of Mass Media. Advances in Economics and Econometrics. Tenth World Congress, vol. 2. Applied Economics. Cambridge, UK. 135-187.
Radhakrishna, R., & Subbarao, K. (1997), India's Public Food Distribution System. World Bank Discussion Paper No. 380, World Bank, Washington, DC.
Reinikka, R., & Svensson, J. (2005), Fighting Corruption to Improve Schooling: Evidence from a Newspaper Campaign in Uganda. Journal of the European Economic Association, 3, 259-67.
Snyder, J. M., & Strömberg, D. (2010), Press Coverage and Political Accountability. Journal of Political Economy, 118 (2), 355-408
Strömberg, D. (2004), Radio's Impact on Public Spending. Quarterly Journal of Economics, 119(1), 189-221.