 |
Versione italiana |
Introduction
In recent times the issue of malnutrition and related consequences on health and economic development has received considerable and justified attention, on a global scale (EU, 2015). It has been widely recognized that macro and micro-nutrient deficiencies can affect the quality of human capital and impact on poverty and economic development. In particular the most detrimental deficiencies tend to be those related to iron, zinc, vitamin A and iodine (Bhutta et al., 2008; Ruel and Alderman, 2013).
Economic access to food through food markets is an important aspect in meeting adequate nutrition (Fao, 2013) but this, in turn, is affected by food price dynamics. Green et.al. (2013) systematically reviewed with meta-regression 136 studies conducted in both developed and developing countries and demonstrated that food consumption in poor countries is more sensitive to price changes than in developed countries. This is because people in developing countries ordinarily spend a much higher proportion of the household income on food. People in developing countries would, therefore, be expected to be most vulnerable to changes in related nutrient intakes in response to increased food prices as they substitute more expensive foods like animal source foods with cheaper less nutrient dense staples. The resulting reduction in dietary diversity could impact negatively on nutrient adequacy, especially with respect to micronutrients intake. Ruel (2003) conducted a review of studies that used dietary diversity methodologies as an indicator of diet quality and found that dietary diversity was positively associated to nutrient adequacy even in poor developing countries.
There is evidence that by means of economic development, developing countries experience changes in food consumption patterns (Vorster et al., 2011) and that the resulting nutrition transition is driven by better economic access to different foods at household level. Some of the changes that take place are positive, e.g. the increasing consumption of animal source protein leading to higher micronutrient intakes, like iron, zinc and vitamin A. On the contrary, other changes in dietary patterns are detrimental to health outcomes. Examples include increased calorie consumption from saturated fat and simple sugars, both associated with increased risk of overweight, obesity and other non-communicable diseases (Ncds) (Vorster et al. 2011). A recent review of studies conducted in several Sub-Saharan African Countries has shown an Increase in terms of overweight individuals in the population of the Countries under consideration (Steyn and Mchiza, 2014). Both micronutrient deficiencies and Ndcs place a significant burden on the national health costs of developing countries making policy intervention an important consideration to mitigate effects of food price volatilities especially in vulnerable countries.
Briggs et al. (2013) conducted a modelling study that explored the effect of a 10% tax on sugar sweetened beverages on obesity in Ireland. The authors reported a small but meaningful effect especially for adults aged 24-34. Although the effect identified by this study was small, the fact that the model only included sugar sweetened beverages should be taken into consideration. Other high sugar containing foods like confectionaries, as well as high fat food items, are also important determinants of obesity. Another study by Claro et al., (2012) found that a tax on sugar-sweetened beverages in Brazil reduced consumption especially for the poor. In the US food and nutrient price elasticity has been reported to have the potential to influence nutrient intake through substitution of foods, as families adjust eating patterns to cope (Miao et al., 2013). Similar effects have been reported in Africa (Abdulai and Aubert, 2004) and Asia (Skoufias et al., 2012). Changes in consumption patterns due to to price changes, regardless of how they are introduced they come about (taxes, local or international price volatility), could impact on nutrient intake and bring about positive or negative consequences. It is not yet clear to what extent such price changes would affect specific nutrient intakes in developing countries. Furthermore, the important role of micronutrients like iron, zinc and vitamin A, protein and energy on health warrant a closer look at the effect of price elasticity on their intakes.
The debate regarding calorie-income relationship is well documented in literature, whereas there is limited research on the relationship between income and key macro and micro nutrients. The literature on income elasticity in relation to calories and nutrients is extensive, and highlights a large heterogeneity in estimates due to differences in research designs, or temporal and spatial dynamics. Many factors tend to influence empirical estimates of income elasticity to nutrients intake: our article aims to systematically review the elasticity of calories, macronutrients and micronutrients to income. In particular we consider a large set of estimates on income elasticity for calories, protein, fat, zinc, iron and vitamin A. The analysis includes studies conducted in developed and developing countries. While previous studies have revised impacts of income on calories intake (e.g. Ogundari and Abdulai, 2013), to the best of our knowledge this is the first review that examines the estimates for income elasticity of calories, micronutrients, and micronutrients on a comparative basis. Moreover we investigate the determinants of the heterogeneity in estimates by means of a rigorous and popular approach: meta-analysis. The information generated may have food pricing policy implications to mitigate possible consequences on nutrient intakes and related health consequences.
Data and methodology
The data employed in the present analysis include numerous studies and estimates on income elasticity. Papers have been collected through most relevant websites for the purposes of the present paper, i.e., Web of Science, Scopus, and Google Scholar. The latter allowed us to cover grey literature (working papers and discussion papers) in order to make sure that that publication bias (i.e. results of published papers systematically different from those of unpublished papers) and the effects of factors such as the journal prestige, and its impact factor can be correctly identified. The studies have been selected according to the presence of information on sample sizes, elasticity, and the associated standard errors or t-values. In total our study includes more than 100 observations (table 1), resulting in a benchmark for future investigations.
We have aggregated the income elasticities for iron, zinc and vitamin A for statistical and epistemological reasons: first the estimates for each micronutrient were less than 20 which renders it unfeasible to apply a meta-analysis; second t-tests revealed that the elasticity of iron and zinc was statistically not different, and the estimates for zinc are have higher mean but very similar variability (i.e. can be assumed to be a mean preserving spread transformation of iron and vitamin A distributions).
A preliminary outcome of the meta-analysis, and indeed a very important step itself, consists in identifying the existence of publication bias. Publication bias may be generated by several factors: preference by authors, reviewers, and editors for statistically significant results to the detriment of studies that report insignificant estimates (Stanley, 2005). The latter, if published, pay the toll of providing no statistically significant results, in terms of collocations in less prestigious journals. Egger et al (1997) suggested to apply funnel asymmetry tests (Fat) to the meta-regression analysis (Mra). The Fat-Mra consists in regressing the effect size of the phenomenon of interest on a measure of the precision of estimates, and other covariates.
We estimate the relationship through a linear regression, and use two sets of covariates, respectively capturing the heterogeneity in estimates, and the publication bias. Following Stanley (2005), our measure of dispersion will be the standard errors of the estimates. Therefore we will test whether the estimates are affected by publication bias.
Table 1 - Studies for the meta-analysis and comparative statistics (elasticities in parenthesis)
Source: Adapted from Santeramo and Khan (2015)
A drawback of the above presented approach is that it suffers for the heteroskedasticity of estimates. We correct this embedded heteroskedasticity by using the inverse of the standard errors, and dividing the dependent variable and the regressors by the standard errors. This procedure will allow to conclude on Publication bias, and the Empirical Effect, i.e. the significativity of income on elasticity.
Finally, we conduct a meta-regression analysis (Mra) to explain the source of heterogeneity in income elasticity of nutrients. The estimation is conducted by mean of Weighted Least Square in order to reduce the effect of publication bias in meta-regression analysis. The weights need to be correlated with the size of the studies, therefore, the inverse of the square root of the standard errors of the estimated effect size are appropriate weights. The approach is similar to that in Ogundari and Adbulai (2013) who weighted by using the inverse of the variance of the standard error of the effect size.
In order to explain the heterogeneity in estimates of income elasticity, we consider several explanatory variables related to the methods of estimation, the number of observations, the location of the study, the prestige of the journal hosting the publication etc. The set of regressors (Table 2) includes the variables “Income”, “Linear”, “Q-Aids”, and “Number of Years”, “Panel”, “Weekly”, “Monthly”, “Rural”, “Africa”, “Asia”, “South America”, “Unpublished Paper”, and “Impact Factor”.
Table 2 - Description of the variables
Q-Aids: Quasi-almost ideal demand system. The model is a workhorse in demand estimation
Source: Adapted from Santeramo and Khan (2015)
Results
Preliminary results suggest that the publication bias may be an issue for the sizes of the effect. The results are robust to heterogeneity in that we have normalized regressors by standard errors. The results show that the bias is detected only for protein and fat (at significance level of 5% or lower): the former shows a negative bias ( ), while the latter has a positive bias ( ). We also found that only for protein and fat the coefficient is statistically significant. We can conclude that the effect of income is relevant to the elasticity of protein and fat reported in the articles and working papers considered in the present analysis. In the other two cases (calories and micronutrients) the effect of income is negligible, which means elasticity tend to be constant, regardless of income level.
Table 3 - Fat-Pet-Mra results
Note: t-stats in parenthesis; +, *, and ** indicate statistical significance at 10, 5 and 1 percent respectively
Source: Adapted from Santeramo and Khan (2015)
Table 4 presents further results that allow to strengthen previous findings. We found further evidence that the influence of income on elasticity is statistically different from zero for protein and fat.
Table 4 - Mst-Mra results
Note: t-stats in parenthesis; +, *, and ** indicate statistical significance at 10, 5 and 1 percent respectively
Source: Adapted from Santeramo and Khan (2015)
Table 5 presents the estimates obtained from Mra. The results show that the elasticity for calories is lower when we adopt household income rather than the expenditure. It is interesting to note that by adopting a Q-Aids model the elasticity tend to be higher, exception made for fat.
The results also provide explanation on how publication bias tends to distort the estimates: the income-elasticity reported in articles published in journals are lower than those reported in working/discussion papers; it is also lower for articles published in journals with impact factor. The variables “Rural” (equals to one if the analysis refers to a rural population) and “Continent” show statistically significant albeit negligible effects. The results on the quality of data are very interesting. Almost in all cases we found that detailed information is associated with larger estimates. For instance, the variable “Panel” is statistically significant and positive. Exception made for fat elasticity. However, in this case albeit the coefficient is negative the significance level is 10%. Similarly, by adopting data at high frequencies (e.g. monthly or weekly) it is likely that higher estimates are obtained. Finally, larger sample sizes may lead to larger or lower estimates, depending on the nutrient considered: elasticity is higher for protein and micro-nutrients, and lower for fat.
Table 5 - Mra results - Weighted regression
Note: t-stats in parenthesis; +, *, and ** indicate statistical significance at 10, 5 and 1 percent respectively
Source: Adapted from Santeramo and Khan (2015)
Conclusions
Food security and nutrition have become central to the policy agendas of governmental and non-governmental organizations due to the consequences that they can generate on health and economic development. Changes in consumption patterns in response to price and income changes could impact on nutrient intake with related positive or negative consequences. A vast empirical literature provides estimates of income elasticity of calories, micro and macro nutrients. However, the elasticities reported in different studies are very heterogeneous. Moreover, due to the empirical and political implications that reported estimates on income elasticity may have, understanding the determinants of the heterogeneity in estimates is of great relevance.
Our meta-analysis found that in the majority of studies calories and proteins are found to be more income-inelastic than fat and micronutrients, which have been found to be more sensitive to income changes. Our meta-analysis found that the influence of income on their elasticity is statistically different from zero. To the extent that adequate child and maternal nutrition have been proved to promote optimal child growth and have positive consequences on economic development (Victora et al., 2008), our results strengthen the importance of implementing policies aimed at improving diet quality and affordability of food. Our results are, therefore, in line with previous studies (Bhutta et al., 2008; Ruel et al., 2013). Moreover, we found a substantial publication bias for sizes of effect that has deserved a deep investigation. For calories, we found that the income-elasticity reported in articles published in journals and particularly in articles published in journals with impact factor tend to be lower with respect to those published in working and discussion papers, thus warning about the reliability of conclusions supported solely by official publications’ estimates. Finally, we found that the quality of data is very important and able to influence estimates.
As a matter of fact, there is a limited number of studies available regarding income-elasticity for protein, fat and micronutrients. Understanding the impact of income changes on nutrients intake remains an important topic deserving further research.
References
-
Egger M., Smith G. D., Schneider M., & Minder C. (1997), Bias in meta-analysis detected by a simple, graphical test. Bmj: British Medical Journal, 315, 629-634
-
European Union. (2015) The role of research in global food and nutrition security. Expo 2015 EU Scientific Steering Committee. Discussion paper
-
Green R., Comelsen L., Dangour A., Turner R., Shankar B., Mazzocchi M. & Smith R.D. (2013), The effect of rising food prices on food consumption: systematic review with meta-regression. Bmj: British Medical Journal, 346
-
Fao, Ifad & Wfp. (2013), The State of Food Insecurity in the World 2013. The multiple dimensions of food security. Rome: Fao
-
Ogundari K., & Abdulai A. (2013), Examining the heterogeneity in calorie–income elasticities: A meta-analysis. Food Policy, 40, 119-128
-
Ruel M. T. (2003), Operationalizing dietary diversity: a review of measurement issues and research priorities. The Journal of Nutrition, 133, 3926S-3926S
-
Ruel M.T., & Alderman H. (2013), Maternal and child nutrition 3. Nutrition-sensitive interventions and programmes: how can they help to accelerate progress in improving maternal and child nutrition?. The lancet, 382, 536-51
-
Santeramo F.G., Khan N. (2015), The income-elasticity of calories, macro and micro nutrients: What is the literature telling us?, Food Research International, 76(4)
-
Santeramo F. G., & Shabnam N. (2015), The income-elasticity of calories, macro-and micro-nutrients: What is the literature telling us?, Food Research International
-
Santeramo F. G. (2015a), Food security composite indices: implications for policy and practice. Development in Practice, 25(4), 594-600
-
Santeramo F. G. (2015b), On the composite indicators for food security: decisions matter! [Link], Food Reviews International [Link]. 31:63-73
-
Stanley T. D. (2005), Beyond publication bias. Journal of Economic Surveys, 19, 309-3
