Based on searches for refereed publications in scholarly journals via EBSCO, JSTOR, and EconLit, on the topic of cigarette consumption, we have selected 18 papers for this review and the results are summarized in Table 2. Important variables effecting cigarette consumption include prices of cigarettes, taxation, per capita income, schooling level, age, expenditure on advertisements and other promotional activities, health indicators, and anti-smoking campaigns.

The main point about the findings in earlier research is that cigarette consumption has all the features of consumption that is addictive, and yet it is useful for policy purposes to see if any particular socio-economic variables have an important bearing on cigarette consumption in the aggregate.

Despite persistent efforts by governmental, inter-governmental and non-governmental agents to significantly reduce cigarette consumption, the ground reality is that consumption has been on the rise. Should there be a further strengthening of the efforts aimed at reducing consumption? If so, are there any significant focus areas: specific groups of nations and age groups, for instance?³ 2Further more, given the strong finding that smoking among pregnant women could cause serious damage to the unborn (Bradford, 2003) 3, is there a likelihood of increasing consumption among women?

The aforementioned issues are addressed in this paper on the basis of cross national multiple regressions with annual per capita cigarette consumption as the dependent variable and the following as the independent variables: percent of population aged above 65 years, female literacy rate and per capita GDP in Purchasing Power Parity (PPP) Dollars. Additionally, to account for possible differential patterns among developed and developing countries, a dummy variable was included in the analysis.

Consumption data were derived from the statistics published by the United States Department of Agriculture (USDA). The data for each country on domestic cigarette consumption in million pieces was obtained from the USDA website. Consumption is estimated as production plus imports minus exports.

Cigarette consumption data are available for 90 countries. There is substantial variation in annual per capita consumption. Minimum annual consumption of cigarettes per capita was found in Rwanda (4 pieces) and the maximum in Malta (3526). The average consumption for all 90 countries put together was 806 pieces. Countries, which are close to the average, are Chile (805), Singapore (836), Malaysia (857), and Indonesia (859).

Asian Development Bank (ADB) Key Indicators and the World Bank World Development Indicators (WDI) were the sources for socio-economic indicators. Wherever there were gaps in those sources, the indicators were taken from Human Development Indicators issued annually by the United Nations Development Programme (UNDP). All data refer to the years 2000/2001 (averages).

Prior to proceeding with multiple regression analysis, simple correlation coefficients are presented (Tables 3 and 4) for a brief review. In the table, APCCC refers to annual per capita cigarette consumption for each country. PCGDP refers to per capita gross domestic product, POP65 to percent of total population aged 65 and over, and FEMLIT to the female literacy rate. Concepts and definitions as well as measures are fully explained in the data sources cited earlier.

Per capita GDP, percentage of population above 65 years, and female literacy are all positively and significantly correlated with per capita cigarette consumption in linear and log linear specifications (Tables 3 and 4). However, an important point to note from the simple correlations is that, the explanatory variables are also correlated among themselves (linear and log linear) positively and significantly. This implies that multiple regressions are likely to be adversely affected due to multi-co-linearity, which typically leads to the explosion of standard errors and the observation of statistical insignificance of the estimated coefficients, an aspect to be re-visited in the following review of the multiple regression results.

Multiple regression equations with linear and log linear specifications and with the four explanatory variables are estimated by Ordinary Least Squares method. The results are presented in Table 5 below. In general, the linear specification is preferred since the adjusted R square is high and all the coefficients are significant.

Between the two linear regressions, A and B, it is clear that in A, multi-co-linearity has taken its toll on the coefficient of per capita GDP which has turned negative. The adjusted R square being about the same between A and B, one could just rely on B. Based on the linear regression B, it turns out that per capita cigarette consumption is affected positively with aging, rising female literacy and industrialization (indicated by the developed country dummy).

The regression results have important implications for the present day developing economies. Per capita cigarette consumption will increase significantly, as those economies aim to raise their overall development status, as on-going fertility declines bring future increases in aged population proportions, and as female literacy increases.

If anything, developing countries are at a threshold of a revolutionary increase in cigarette consumption. These economies offer enormous marketing opportunities for the cigarette manufactures. Unless doubling and trebling of the efforts to control cigarette consumption are in place, there is no way to restrain consumption.