Chronic child malnutrition and anemia are among the main risk factors for child development across the developing world (Walker et al., 2007). Childhood anemia is associated with serious consequences that include growth retardation, impaired motor and cognitive development, and increased morbidity and mortality (Grantham-McGregor et al., 2001; Ramakrishnan, 2008). Children with malnutrition problems are nine times more likely to die than those who do not suffer this condition (Black, el al. 2013).

In Peru, 14% of children under 5 years of age are chronically malnourished and 36% of children between 6 and 36 months of age show some degree of anemia (INEI, 2015). There are also significant disparities according to their geographical domain. While the prevalence of chronic malnutrition in urban areas is 9%, in rural areas this figure reaches 27%. In addition, 44% of children in rural areas suffer anemia compared to 34% in urban areas. In this sense, chronic child malnutrition and anemia causes are related to families’ access to an adequate food intake, basic public services and, more in general, to their socioeconomic status. Therefore, a relevant policy question concerns which improvements can be expected in child malnutrition and anemia if shifts in these determinants are produced.

In this paper we seek to estimate the improvements that can be expected in child malnutrition and anemia in Peru if Sustainable Development Goals (SDGs) related health determinants reach a set of target values consistent with the targets proposed for the SDGs by the United Nations. We situate the determinants of child malnutrition and anemia within the scope of the SDGs for two reasons. First, SDGs provide a sufficiently ample array of social variables. Second, it adds policy relevance because it will focus the analysis on variables that the policy maker has already committed to shift.

This research contributes to the literature in two ways. To begin with, it is the first study to offer improvement scenarios for SDG-related health outcomes based on shifts produced in other SDG-related variables. SDG indicators comprise outcome variables but also variables that can be linked directly and indirectly to these outcomes. Therefore, it is more informative for a policymaker committed to achieving the SDGs if improvements in SDG-related outcome variables are assessed by producing improvements in SDG-related determinants, especially if these determinants have a direct link with policy action (e.g. they refer to access to a basic public service).

Outcome variables such as child malnutrition and anemia have a direct connection with children’s and families’ wellbeing but cannot be directly influenced by policy (one cannot treat children with less anemia). By linking progress in health outcomes to improvements in their SDG-related determinants, the policymaker will have a better idea of what to expect in terms of improved wellbeing if he/she fulfills his/her commitments in terms of variables that have a more direct connection with policy.
Efforts have been made to measure levels and progress in achieving health-related SDGs and project its attainment by 2030 (GBD 2016 SDG Collaborators, 2017). Contributions have focused on expanding the array of indicators available across countries and on estimating improvement scenarios for individual indicators based on historic trends. To the best of our knowledge, no previous study has estimated improvement scenarios for SDG health outcomes based on their relation with other SDGs that can function as health outcome determinants.

The second contribution is related to the methodology employed to estimate the improvement scenarios for chronic child malnutrition and anemia. SDG-related health determinants comprise both health inputs and health input determinants. Inputs are variables that have a direct effect on the outcome (e.g. food intake), while input determinants have an indirect effect as they are mediated by inputs (e.g. maternal education). Inputs and input determinants cannot be treated the same way in empirical work. Failure to recognize this distinction can lead to a misinterpretation of the improvement scenarios. In this sense, if one seeks to estimate the improvement in child health produced by a shift in maternal education, one has to consider the results of a regression of child health on input determinants only. Otherwise, regressing child health on inputs (e.g. food intake) and input determinants (e.g. maternal education) will provide the expected improvement in child health if the unobserved inputs reach a level consistent with input determinants reaching a certain target value, mediating the effect through the unobserved inputs.
In the analysis we distinguish between three different types of empirical specifications. The first one is a production function and includes only health inputs. The second specification is a demand function and contains only health input determinants. The third is a hybrid production function and includes both health inputs and input determinants. Each specification requires particular assumptions to produce the estimate of interest, which is the expected improvement in the health outcome (chronic child malnutrition or anemia) after SDG-related determinants reach their target values. We will use the insights of a simple economic model describing how families’ decisions translate into particular levels of health inputs to explain these assumptions and choose the most appropriate empirical specification. We believe this methodology could be used in future efforts seeking to estimate improvement scenarios in SDG-related outcomes produced by shifts in SDG-related determinants.

Consistent with the SDGs and based on the information available in the Peruvian National Health and Demographic Survey (ENDES, 2015), the input variables considered reflect adequate food intake (related to SDG 2, target 2.1), adequate dwelling conditions (SDG 6, targets 6.1, 6.2; SDG 7, target 7.1; SDG 11, target 11.1), and access to health services (SDG 3, target 3.8). Target values are set so as to reflect universal access to these basic goods and services. SDG-related input determinants correspond to household wealth (related to SDG1, target 1.2) and maternal education (SDG 4, target 4.3). The target value for household wealth was defined to reflect a shift in the ENDES 2015 household wealth index distribution such that all households have a score equal or greater than the 2015 median value. The target value for maternal education was defined considering the average proportion of women with secondary and tertiary education in the OECD member countries.

Our main findings can be summarized as follows. We can expect a significant reduction of 8.9 percentage points (from 14.6% to 5.7%) in chronic child malnutrition in Peru if all the SDG-related determinants reach their targets. Around 63% of this improvement (5.6 percentage points) is achieved through the effect of observable inputs. The remaining 3.3 percentage points can be attributed to the shift in unobservable inputs produced by input determinants reaching their targets. The potential for improvement is much more significant in the rural domain. Simulations reveal that we can expect a reduction of 21.6 percentage points (from 27.4% to 5.8%) in chronic child malnutrition in rural Peru if all the SDG-related determinants reach their targets. Around half of this improvement (11.2 percentage points) is produced by the shift in observable inputs.
In a similar way, if all SDG-related health determinants reach their targets, we can also expect an important reduction of 15.8 percentage points in childhood anemia in Peru (from 36.4% to 20.6%). Around 65% of this reduction can be achieved through the effect of observable inputs. As in the case of malnutrition, rural areas show a greater potential for improvement. In particular, we can expect a reduction of 21.5 percentage points (from 43.8% down to 21.5%) in childhood anemia in rural Peru if all the SDG-related determinants reach their targets. More than half (53%) of this improvement is produced by observable inputs.

Further discussion and research should acknowledge that health outcomes such as child chronic malnutrition and anemia could respond to multiple causes. In this sense, data limitations could prevent one from fully characterizing these causes and, therefore, some inputs will inevitably remain unobserved. This poses an important challenge for an empirical exercise aiming at estimating improvement scenarios that can be related to shifts in policy relevant determinants for two reasons. First, unobservable inputs can act as confounders when estimating the relation between observed inputs and health outcomes. Secondly, and finally, not all observed determinants are equally amenable to policy action. In this sense, it will be of little policy relevance if most of the improvement produced by shifts in observable determinants is related to variables that do not have a direct connection with policy action; nonetheless, this analysis serves as an example of compatibility between policy-related determinants and significant empirical inputs.