Chatham House report
Published 8 July 2020
Updated 17 December 2020
ISBN: 978 1 78413 404 4
Malnutrition acts as a brake on the development of individuals, communities and economies around the world. This report is the first of its kind to reveal the hidden costs of malnutrition for business, and the extent to which these costs are recognized and addressed by multinational companies (MNCs).
Here we explain the methodology and data sources used in the modelling analysis to answer the question: ‘What would happen to business output in the short term if workers achieved the physical state associated with good nutritional outcomes?’
We start in Section 1.1 with a review of the human capital approach to estimating worker productivity, followed by a review of the literature on the direct impact of malnutrition on worker productivity. Subsequent sub-sections review the evidence on four specific types of malnutrition: underweight, obesity, anaemia and childhood malnutrition.
Section 1.2 briefly outlines the theoretical approach behind the modelling analysis, before Sections 1.3 to 1.6 describe the steps involved in generating the estimates used to inform this report’s conclusions. As Figure 12 summarizes, there were five stages to the modelling analysis. Below, we explain the data used and our modelling approach.
This analysis assumes that malnutrition reduces the capacity of the labour force through compromising the quality of human capital. Mincer’s146 model of human capital expresses wages as a function of worker characteristics.147 The total remuneration of a worker is viewed as a base wage (the marginal productivity of a benchmark worker) plus a wage premium associated with each worker characteristic that raises human capital.148 A typical Mincer wage equation would be:
where Wi is the total wage, Wo is the benchmark wage, Si is education, Ei is labour market experience, Hi is a vector of physical and mental health variables, and Xi is a vector of control variables denoting labour market and demographic characteristics. This leads to the widely estimated equation:149
where εi is an error term.
Malnutrition augments numerous terms within the human capital model, highlighting its relevance to the issue of labour productivity. Malnutrition impacts the education (Si) an individual obtains, as children with nutritional difficulties tend to enrol in school later than other children, progress more slowly across grades and attain lower levels of scholarly achievement. Furthermore, malnutrition is the leading cause of ill-health and disability (Hi), which also contributes to lower levels of labour market experience (Ei).
An extensive literature demonstrates that investments in human capital in labour markets produce substantial returns. The literature, especially in the low-income, lower-middle-income and upper-middle-income settings relevant to this analysis, has largely focused on the returns from education.150 Papers which include health outcomes151 typically find that the returns from physical development and capacity are smaller than those from education. For example, Thomas and Strauss152 find that among male workers in urban Brazil, a literate man earns 50 per cent more than an illiterate man, while a 1 per cent increase in the height of a man is associated with a 2.4 per cent increase in wages. Holding all other factors constant, an illiterate male would have to be 30 cm taller than a literate male to earn the same wage. Whereas this discrepancy is significant, it is likely that workers in low-income countries have fewer opportunities to reap the returns from education in knowledge-intensive sectors than do workers in middle-income countries such as Brazil. In low-income countries, workers are expected to benefit more from physical development and strength. Nonetheless, these results indicate that the impact of malnutrition on worker productivity by suppressing educational attainment can be significant.
As explained above, this study focuses primarily on the direct impact of malnutrition on businesses in terms of reducing workers’ physical and cognitive capacity. Physical work capacity is dependent on an individual’s maximal oxygen uptake: the higher the oxygen intake, the greater the person’s capacity to convert energy into effort.153 Several forms of malnutrition affect an individual’s maximal oxygen uptake, and therefore physical capacity:
Obesity affects productivity through a separate biological mechanism. Instead of impairing maximal oxygen uptake, obesity increases the risk of various diseases such as cancer, diabetes, depression and arthritis, which are assumed to adversely affect labour productivity. Furthermore, obese people can experience greater difficulties with physical tasks and with completing tasks on time.154
The sections below introduce the literature used to estimate the impact of four expressions of malnutrition on labour productivity: underweight, obesity, anaemia and the experience of malnutrition in childhood. The model parameters, discussed in Chapter 2, are drawn directly from this literature review.
In low-income or middle-income countries, increased BMI has a significant positive impact on output and wages
Underweight reduces the physical and cognitive capacity of workers, which negatively affects labour productivity. There are a number of ways to measure whether a person is underweight or experiencing chronic hunger (which can give rise to underweight). These include both input indicators such as caloric intake and outcome indicators such as BMI. Studies have shown that both input and outcome indicators of underweight and chronic hunger are associated with lower physical and cognitive work capacity.155 Our model utilizes the latter set of indicators. Measuring the impact of underweight on labour productivity is challenging due to simultaneity issues: underweight reduces worker productivity, resulting in low wages which in turn perpetuate chronic hunger – this is known as the efficiency wage hypothesis. Table 3 summarizes how studies have used various approaches to overcome these endogeneity issues. The studies find that in low-income or middle-income countries, increased BMI has a significant positive impact on output and wages. This result is found for both men and women, and in both urban and rural contexts. The impact of being underweight seems to be more severe among manual workers and those with lower educational attainment, who are more likely to work in physically demanding roles.
Table 3 includes four papers156 which estimate the specific impact of being underweight on a person’s labour productivity. These papers are used to establish appropriate coefficients for use in the model.
|
Study |
Context |
Dependent variable |
Explanatory variables |
Methods |
Results |
|---|---|---|---|---|---|
|
Carrillo and Charris157 |
Male and female workers, Brazil |
Hourly earnings (calculated by monthly labour income divided by hours worked) and sectoral choice (formal, informal) |
BMI, individual characteristics (gender, ethnicity, urban/rural), education |
Cross-sectional data, instrumental variable approach (sibling BMI) |
A 1% increase in BMI is associated with a 0.5% increase in earnings; this effect is larger for women and for workers living in urban areas. |
|
Lafave and Thomas158 |
Male workers, central Java (rural Indonesia) |
Hourly earnings (calculated as total earnings during the previous four months divided by hours worked), occupational choice |
Height, cognition, education, health (BMI, blood pressure, self-reported ability to run 1 km) |
Panel data, fixed effects |
A 1% increase in BMI is associated with a 1.2% increase in hourly earnings, holding other explanatory variables constant. |
|
Kedir159 |
Male and female workers, urban Ethiopia |
Monthly wage |
Education, height, BMI, individual characteristics (age, experience, location) |
Panel data, instrumental variable approach (food prices, household size) |
A 1% increase in BMI increases monthly wage by 3.7%; this effect is larger for male workers and people with lower education. |
|
Colchero and Bishai160 |
Working mothers,161 metropolitan Cebu (urban Philippines) |
Hourly earnings |
Nutrition (underweight, normal, overweight and obese), source of income (wages, piece rate, self-employed, multiple), education, age, breastfeeding |
Panel data, fixed effects, instrumental variable (lagged BMI) |
Hourly earnings were 8.6% higher among healthy-weight women than in underweight women. The impact of BMI was greater for self-employed women and those with multiple occupations. |
|
Luo and Zhang162 |
Male and female workers, China |
Employment status, monthly wage |
Nutrition indicators (BMI, BMI2, underweight, overweight, obese), individual characteristics (age, hukou status, ethnicity, educational attainment, marital status, experience), self-reported health status, occupation |
OLS regression, fixed effects |
Non-linear impacts of BMI on employment and wages. Being underweight does not significantly affect male employment or wages, but significantly affects female wages (reduction of 40% relative to those with healthy BMI). The authors suggest discrimination is driving the results. |
|
Yimer and Fantaw163 |
Male and female workers, urban Ethiopia |
Monthly wage |
Height, BMI, schooling, experience |
Panel data, instrumental variable quantile regression (past BMI values) |
The log wage elasticity with respect to BMI is 0.059. BMI is found to have a statistically significant impact on wages at all income quintiles and is a stronger determinant of female wages and the wages of younger cohorts. The effect of BMI was stronger for those on low incomes. |
|
Shimokawa164 |
Male and female workers, China |
Monthly wage |
Nutrition (underweight, overweight, obesity), height, education, individual characteristics (age, marital status, household size, breastfeeding, region) |
Panel data, multiple approaches used, both parametric (including fixed effects and instrumental variable regressions), and semi-parametric |
Wage penalty for being underweight or obese. The wage penalty is larger and more widespread among men than among women in China, owing to the prevalence of manual labour. |
|
Dinda, Gangopadhyay and Chattopadhyay165 |
Male coalmine workers, India |
Monthly wage |
Height, BMI, environmental conditions, individual characteristics (experience, age) |
Cross-sectional data, OLS |
Underweight workers earn 2% less than the reference standard wage. |
|
Schultz166 |
Male and female workers, Côte d’Ivoire and Ghana |
Hourly wage |
Height, BMI, education, migration status |
Instrumental variable approach (community health infrastructure, food prices, parental education) |
A unit increase in BMI is associated with a 9% increase in women’s wages in both Côte d’Ivoire and Ghana. A unit of BMI increases men’s wages by 15% in Côte d’Ivoire and by 7% in Ghana. Education and BMI appear somewhat substitutable, particularly for women. |
|
Croppenstedt and Muller167 |
Male agricultural workers, rural Ethiopia |
Agricultural output, daily wage |
Agricultural inputs (such as land area and quality), water availability, nutrition of household head (weight-for-height, BMI), education |
Cross-sectional data, instrumental variable approach |
The male wage elasticity with respect to the BMI is 3.04, which means that an increase of one standard deviation would increase the wage by 26%. |
|
Thomas and Strauss168 |
Male and female workers, urban south and northeast Brazil |
Hourly earnings (calculated as annual earnings divided by hours worked) |
Height, BMI, calorie intake, protein intake |
Cross-sectional data, instrumental variable approach (food prices) |
A 1% increase in BMI is associated with a 2.2% increase in male wages; this effect is larger among the self-employed and those with less education. The effect of BMI is smaller and not statistically significant for women, except for the least educated women. |
Source: Vivid Economics.
Relatively few studies in developing countries have considered the implications of obesity for labour productivity. In developed countries, where obesity is more common, obesity has been found to have a significantly negative impact on labour market outcomes; some of this is attributable to discrimination.169 For example, a 2016 OECD study finds that across 14 European countries in 2013, 59 per cent of obese people aged 50–59 were employed, compared to 72 per cent of people who were not obese. Among the employed, obesity is found to increase the likelihood of worker absence and to lower labour productivity. Across these issues, obese women experience greater difficulties than obese men. The only studies in developing countries to have considered the labour market implications of obesity are Luo and Zhang170 and Shimokawa.171 These studies both look at the impact of obesity in China during the country’s transition period to more ‘Western’ diets, when there was a rapid increase in the prevalence of overweight and obesity while the population was still experiencing food insecurity. The two papers find that obesity has negative labour market impacts, consistent with the developed-country literature.
Studies show that anaemia reduces worker productivity, even when workers are consuming sufficient calories.172 Most studies investigating the impact of anaemia have been limited to laboratory experiments or randomized controlled trials, which have estimated the impact of iron supplementation on labour productivity. For example, Edgerton et al.173 find that the productivity of workers on a tea plantation in Sri Lanka increased in response to iron supplementation. Horton and Ross174 review these studies and conclude that iron therapy in anaemic adults is associated with a 5 per cent increase in labour productivity among light manual workers and a 17 per cent increase in productivity among heavy manual workers. These estimates are in line with Weinberger,175 who finds that wages would be 5–17.3 per cent higher if households achieved recommended levels of iron intake.
Macronutrient and micronutrient deficiencies at different stages of childhood can inhibit growth, development and cognitive function. Nutritional deficiencies at any stage of life can have implications for productivity capacity and cognition, but are particularly impactful during the early formative stages of growth. Undernutrition also increases child morbidity and lowers educational attainment. Anaemia (iron or vitamin B12 deficiency) and zinc deficiency can inhibit child growth and development, while vitamin A deficiency and iodine deficiency disorder can increase the risk of severe diseases and child mortality.176 The costs of child undernutrition on society include poorer health outcomes and the associated burden on society, reduced educational attainment, and lower physical and cognitive capacity.177
This model analysis uses stunting as a proxy for childhood nutritional deficiencies to represent the most severe impacts of hunger and micronutrient deficiencies in childhood. Stunting is a consequence of maternal malnutrition, undernutrition during infancy and impaired absorption of nutrients. Stunting serves as a proxy for all forms of nutritional deficiencies in childhood because it reflects the strongest impacts in terms of impaired child development, and because its inclusion in the analysis resolves the difficulty of estimating the timing of undernutrition by nutrient during different stages of development.178 Childhood stunting is expected to reduce educational attainment, physical stature and cognitive capacity, impacting future labour productivity.179
Due to the static nature of the model, the analysis focuses on the impact of childhood malnutrition on physical development. In line with the Cost of Hunger in Africa studies180 and the Nutrition PROFILES guidelines,181 this modelling analysis draws on a study in the Philippines in which the wages earned by sugar cane workers were 1.38 per cent higher for every 1 per cent increase in their height.182,183 We interpret the study as an estimate of the impact of malnutrition on labour productivity through the mechanism of reduced physical development.
Childhood malnutrition has been shown to have various negative impacts on cognitive development, educational attainment and adult physical capacity. Childhood malnutrition is assumed to impair physical development, compromising the labour productivity of manual workers. It can also have an indirect impact on business through reducing the educational attainment of the workforce. Malnourished children tend to enrol in school later, progress more slowly through grades and attain lower levels of scholarly achievement. Using stunting as a proxy for childhood malnutrition, Galasso and Wagstaff184 use a meta-analysis to show that stunted children attain 1.6 fewer years of education.
This box is intended to provide an indication of how much larger the results would be if this indirect effect were additionally considered. The labour productivity implications of reduced educational attainment are context-dependent, as high-skilled labour is complementary to businesses’ use of more technologically advanced machinery.185 An additional year of education has been found to increase wages by 5–11 per cent.186 Assuming a tight link between wages and productivity, this would suggest that childhood malnutrition reduces the effectiveness of labour by 8–18 per cent. As such, the impacts of education losses from childhood malnutrition on business are expected to outweigh the physical impacts by a factor of 1.6–3.5.
The physical impact of childhood malnutrition was estimated to impose a cost to business equivalent to 0.4 per cent of GDP annually across the 17 countries for which stunting is modelled. If the education-related indirect impacts are included, this could increase to as much as 1.8 per cent of GDP.
Malnutrition is assumed to impact business through compromising the quality of human capital. Figure 13 summarizes the impact channel framework used to model the cost that malnutrition poses to business. The model focuses on the relationship between adult malnutrition and physical development, and on the physical impacts of stunting experienced in childhood. As a result of compromised physical development and reduced physical and cognitive capacity, malnutrition leads to both presenteeism (reduced productivity at work) and absenteeism (physical absence from the workplace). These effects combine to reduce the overall capacity of the workforce, reducing output and negatively affecting businesses.
The model focuses on a narrow aspect of the malnutrition problem. It answers the simple question: ‘What would happen to business output in the short term if workers achieved the physical state associated with good nutritional outcomes?’ Given the broad-ranging and complex relationships between nutrition, productivity and the economy, this model does not take into account the following impacts on business (see also Figure 14):
Table 4 summarizes the mechanisms by which malnutrition affects business and which of these are included in the modelling analysis.
|
Population |
Mechanism |
How does this mechanism impact business? |
Incorporated into the modelling analysis? |
|---|---|---|---|
|
Workforce (current) |
Reduced physical and cognitive capacity |
Underweight, hidden hunger and obesity reduce the physical and cognitive capacity of the workforce, resulting in presenteeism. |
The impact of underweight on the labour productivity of the workforce is a core aspect of the model. |
|
The impact of obesity on the labour productivity of the workforce is a core aspect of the model. |
|||
|
Data were only available to calculate the prevalence of anaemia in five countries; therefore this is considered an extension to the model and the results cannot be extrapolated. |
|||
|
Workforce (childhood) |
Physical development |
Childhood malnutrition compromises the physical development and strength of the adult workforce, reducing labour productivity. |
The impact is modelled for 17 countries, as proxied by adult short stature. However, the unclear business implications mean this is considered an extension to the core model. |
|
Workforce (childhood) |
Cognitive development and educational attainment |
Childhood malnutrition compromises the cognitive development of children and results in lower educational attainment, reducing the productivity of the workforce. |
Not modelled. This mechanism would result in individuals choosing/following different occupations. Given the static nature of the model, this self-selection could not be modelled. |
|
Workforce (current and childhood) |
Increased risk of morbidity and mortality |
Malnutrition (whether experienced in the past or present) increases an individual’s risk of morbidity and mortality, resulting in increased absenteeism and reduced labour supply. |
Not modelled due to insufficient evidence in the literature. |
|
Total population |
Reduced disposable income |
Lower labour productivity due to malnutrition will suppress wages (this is known as the poverty-malnutrition trap). Furthermore, malnutrition results in higher private healthcare costs. Consequently, malnutrition leads to lower disposable incomes, and lower demand for consumer goods. |
Not modelled due to the static nature of the model. |
|
Total population |
Compromised enabling environment |
Malnutrition reduces the fiscal space available to governments. It increases public healthcare spending and lowers tax revenue. This restricts government ability to invest in business-friendly priorities. |
Not modelled due to the static nature of the model. |
|
Total population |
Supply chain impacts |
Malnutrition in upstream and downstream sectors may reduce the supply of a business’s inputs or reduce demand for its outputs. |
Not modelled due to the static nature of the model. |
|
Total population |
Dynamic implications |
Overcoming malnutrition would support structural economic transformation and influence rural–urban migration patterns, the implications of which for an individual business are unclear. |
Not modelled due to the static nature of the model. |
Source: Vivid Economics.
The model examines a stylized version of the economy in which the effects of improved nutrition on the structure of the economy, and of the labour force, are not considered. The modelling analysis presents a static and partial equilibrium view of business performance. The model assumes that the structure of both the economy and the labour force are independent of malnutrition, holding constant the relative sizes of the sectors of the economy, the size of the labour force, labour force participation rates, business decisions to invest in capital goods, and the rate of substitution between labour and capital by sector. Reducing malnutrition is likely to support economic structural transformation. There is evidence in the literature that people self-select into occupations based on their nutritional status.188 For example, a healthier and more nutritionally secure worker may select a more physically or cognitively demanding occupation if there is a greater opportunity for higher income. Furthermore, overcoming malnutrition in childhood would result in a better-educated and more cognitively developed workforce which would work in more highly skilled occupations.189
A further factor not included in the modelling analysis is the impact of malnutrition on peace and stability. Extreme malnutrition, where people lack access to even basic food supplies, can be a cause of forced migration.190 Hunger has a destabilizing effect on societies, and there are historical examples of protests and political turmoil following periods of famine or rising food prices.191 Overcoming malnutrition could help contribute to peace and stability, significantly improving the operating environment for businesses.
Having described the theoretical approach to the modelling analysis, in the following sections we explain in detail the first four stages of the approach.
The first stage of the modelling analysis involves estimating the prevalence of malnutrition by economic sector. Using household survey data, the prevalence of underweight and obesity is estimated for 13 economic sectors across 19 low-, lower-middle- and upper-middle-income countries (with the prevalence of anaemia and short adult stature estimated for smaller subsets of the same countries, across the 13 economic sectors).192 The measure/proxy used for each form of malnutrition modelled is as follows:
The analysis is based on malnutrition outcome indicators, as opposed to nutrition input indicators such as calorie or micronutrient intake. Firstly, the selected outcome indicators are convenient to measure and consequently have been extensively used in the literature, indicating widespread acceptance.197,198 Secondly, unlike input indicators, outcome indicators do not distinguish between different causes of malnutrition, which is important as WASH, education and health services can greatly exacerbate the health issues caused by inadequate nutritional intake.199 Finally, these indicators allowed us to use the USAID-funded Demographic and Health Surveys (DHS) programme as the basis of the model; the DHS programme reports individual-level data consistently across a large number of developing countries.
For the countries covered by our model, DHS data are used to estimate the prevalence of malnutrition by occupation and gender.200 The DHS programme conducts nationally representative household surveys collecting data in the areas of population, health and nutrition. These surveys are administered consistently across 92 developing countries. The DHS programme collects data both on the employment occupation of individuals and on a range of biomarkers and anthropometry associated with nutritional status. Our analysis identifies malnourished survey respondents according to World Health Organization (WHO) guidelines:
Box 12 summarizes how the DHS data are used in the modelling exercise.
Surveys under the DHS programme have large sample sizes (usually between 5,000 and 30,000 households) and are typically conducted every five years.
Our modelling exercise required three types of variables available in the DHS:
The prevalence of malnutrition by occupation, extracted from the DHS data, is paired with International Labour Organization (ILO) data in order to estimate the prevalence of malnutrition by sector. We also use recorded occupation to identify which workers are undertaking light or heavy manual labour, using the most detailed occupation classification recorded. We first estimate the prevalence of malnutrition by type and occupation, and then map the prevalence to sectors’ occupational composition as illustrated in Figure 15. ILOSTAT country data, detailed in Box 13, are used to calculate the distribution of occupations across sectors.
ILOSTAT data are compiled from national labour force surveys. In this project three datasets are used:
Employment estimates are scaled to 2017. We use the ILO’s modelled employment estimates to scale up or down the employment estimates for the most recent year of observed employment data. For example, if the most recent year of employment figures for a country is 2015, we use the percentage change between the 2015 and 2017 modelled estimates to scale the employment figures. This scaling method is applied to any employment figures not reported in 2017. This assists in adjusting the size and structure of the labour force across countries.
For countries where weight and height data are collected only among female survey respondents, we estimate the male prevalence of malnutrition by occupation using national-level estimates of underweight and obesity. The WHO Global Health Observatory reports the share of the population which is underweight or obese at a country level, disaggregated by gender. We estimate the prevalence in males by multiplying the prevalence in females by occupation by the male-to-female prevalence ratio at the country level. We do not use this method for estimating male anaemia prevalence, due to the lack of national-level data on male anaemia; for anaemia estimates, we only use countries where both male and female biometrics are reported.
The next stage of the analysis involves identifying the impact of malnutrition on the labour productivity of the individual. The model focuses on two core forms of malnutrition: underweight and obesity. For these forms of malnutrition, low, central and high impact scenarios have been developed to create a plausible range for output results, shown in Table 5. The other two forms of malnutrition for which sectoral prevalence is estimated are modelled as extensions to the core model. In the case of anaemia, this is due to data availability, as only five of the DHS surveys used include measures of haemoglobin levels for both male and female respondents, and we are unable to use national-level estimates for males since data on anaemia in men are rarely collected. In the case of childhood stunting, this is due to interpretation considerations. As businesses cannot affect the historical nutrition experiences of their workforces, the impact of childhood stunting on current business performance is not considered a core model result.
As businesses cannot affect the historical nutrition experiences of their workforces, the impact of childhood stunting on current business performance is not considered a core model result
Coefficients of labour productivity loss from the literature define the parameters for modelling the impact of malnutrition. Table 5 summarizes the coefficients used. As evident in Table 3, much of the existing literature estimates the impact of malnutrition on wages rather than on labour productivity directly. In line with Van Biesebroeck,206 our analysis assumes that income loss for the individual is directly proportional to the individual’s productivity loss. Bias against workers whose malnutrition is evident (for example, because they are short) may cause this assumption to fail. The literature reviewed did not concretely determine whether bias or discrimination was driving the results. For example, several studies find that physical attributes have greater impacts on the earnings of the self-employed, suggesting limited employer discrimination.207 However, Luo and Zhang208 attribute the negative impact of obesity on female employment to discrimination, on the basis that self-reported health status and self-confidence do not fully explain their findings. It is unclear, therefore, how robust this approach is.
|
Malnutrition |
Gender |
Low impact |
Central impact |
High impact |
Notes |
|---|---|---|---|---|---|
|
Underweight |
Male |
0.94 |
0.74 |
0.74 |
Values adapted from Shimokawa.209 The central impact value is identical to the high impact value. This is to reflect the literature: some studies find the impact of underweight to be greater for men;210 others find the impact is greater for women.211 Given this, it was decided that the central impact coefficients should be the same for men and women. |
|
Underweight |
Female |
0.90 |
0.74 |
0.59 |
Values adapted from Shimokawa.212 |
|
Obesity |
Male |
0.95 |
0.89 |
0.83 |
|
|
Obesity |
Female |
0.93 |
0.67 |
0.41 |
Values adapted from Shimokawa.215 |
Note: The coefficients represent labour force effectiveness relative to no impact of malnutrition. A coefficient of 0.94 represents a person working at 94 per cent capacity as a result of malnutrition. Values closer to 1 suggest a lower impact from malnutrition, and values further below 1 suggest a greater impact from malnutrition.
This study is based on best available evidence studying the relationship between nutrition outcomes and labour productivity. Where possible, we use a range of estimates from multiple sources, or from a single source where evidence is extremely limited. Accordingly, we note that there are limitations to the interpretation of these coefficients, and that this is an area for future research to explore further the impacts of poor nutrition on labour productivity.
Source: Vivid Economics.
The next stage in the analysis models the impact of malnutrition on sector output. The prevalence of malnutrition and the impact of malnutrition on labour productivity are combined in a simple multisectoral economic model which estimates the impact of malnutrition on sectoral gross value added (GVA). Figure 16 summarizes the economic model used.
The model estimates the impact of malnutrition as the difference between current sectoral output and potential sectoral output without the reduction in workforce effectiveness from malnutrition. Malnutrition is assumed to impact the aISIC, c term: the labour supply is less effective due to malnutrition. The sector-specific term aISIC, c is calculated based on the prevalence of malnutrition in that sector and the impact of malnutrition on labour productivity. To illustrate this approach, consider a hypothetical case in which underweight reduces the labour productivity of an agricultural worker by 30 per cent and 10 per cent of agricultural workers experience underweight. Across the agricultural sector in this scenario, underweight reduces labour productivity by 3 per cent, such that aISIC, c is 97 per cent. The losses from malnutrition are calculated as the difference between current sector GVA (where malnutrition among the workforce means aISIC, c < 1) and potential GVA in the ‘no malnutrition’ scenario (aISIC, c =1).
Table 6 summarizes the sources of data for the parameters used to calibrate the model. These data sources are:
Nutritional deficiencies are assumed to be additive in impact. The literature used to determine the coefficients typically aims to estimate the impact of malnutrition on labour productivity holding other explanatory variables constant. For example, the impact of BMI on wages is estimated holding height constant, and the impact of anaemia on productivity is estimated holding calorie consumption constant. However, the literature does not investigate the additionality or interaction between forms of malnutrition, and it is unclear whether the presence of multiple forms of malnutrition would amplify or reduce the observed impact of specific forms of malnutrition on productivity. It is unclear, therefore, how conservative our approach is.
Malnutrition among workers is expected to contribute to absence from work due to illness, and there is a growing body of evidence that shows that poor health in children imposes costs on business through increasing parental absenteeism
Malnutrition is expected to affect both presenteeism and absenteeism. The literature reviewed has only estimated the impact of malnutrition on output, hourly wages or monthly wages. While monthly wages may take into account some absenteeism impacts, it is reasonable to conclude that the literature has focused on presenteeism. Malnutrition among workers is expected to contribute to absence from work due to illness, and there is a growing body of evidence that shows that poor health in children (which can be caused by malnutrition) imposes costs on business through increasing parental absenteeism.218 The modelling exercise only takes into consideration the impacts of presenteeism on the current workforce.
Given the static, partial equilibrium approach, only labour productivity is assumed to vary with the prevalence of undernutrition. All other inputs to the model are not sensitive to undernutrition. Specifically, employment for each sector is based on current shares of employment by sector for each country; there is no automatic rebalancing between sectors: i.e. the model does not include any cross-sectoral substitution of labour or capital stock. The limitations of this approach are expanded on in the description of the theoretical approach (Section 1.2).
Sensitivity analysis is used to test the results of the modelling and ensure robustness. Four scenarios were developed, picking up on themes in the literature which identified the largest impacts of malnutrition on the workforce. Table 7 summarizes the changes in coefficients used in these four scenarios.
|
Malnutrition |
Gender |
Type of worker |
Scenario |
Notes |
|||
|---|---|---|---|---|---|---|---|
|
1 |
2 |
3 |
4 |
||||
|
Anaemia |
n/a |
Non-manual |
1.00 |
– |
– |
– |
Taken from Horton and Ross.222 |
|
n/a |
Light manual |
0.95 |
– |
– |
– |
||
|
n/a |
Heavy manual |
0.83 |
– |
– |
– |
||
|
Childhood malnutrition (adult short stature) |
n/a |
Non-manual |
– |
1.00 |
– |
– |
Adapted from Haddad and Bouis,223 following the approach used in the Cost of Hunger in Africa studies224 and the Nutrition PROFILES guidelines.225 |
|
n/a |
Manual |
– |
0.94 |
– |
– |
||
|
Severe chronic hunger |
Male |
n/a |
– |
– |
0.61 |
– |
The coefficients are adapted using Shimokawa,226 which shows that the impact of being severely underweight is 60% worse than being underweight for men and 80% worse for women. The coefficients are adapted on the assumption that 10% of people experiencing chronic hunger will experience severe chronic hunger. |
|
Female |
n/a |
– |
– |
0.57 |
– |
||
|
Chronic hunger |
Male |
n/a |
– |
– |
0.75 |
– |
|
|
Female |
n/a |
– |
– |
0.76 |
– |
||
|
Underweight |
Both |
Non-manual |
– |
– |
– |
1.00 |
The coefficients are adapted using Kedir.227 The findings show that physical attributes only determine the wages of individuals with primary-level schooling or less. They argue that these individuals are manual workers and that those with above primary school education are non-manual workers. The coefficients are adapted on the assumption that in low-income and lower-middle income countries 83% of men and 84% of women work in manual operations. |
|
Both |
Manual |
– |
– |
– |
0.69 |
||
Note: Childhood malnutrition is proxied by childhood stunting, which in turn is proxied by adult short stature, as described in previous sections, while severe chronic hunger is proxied by BMI <17 kg/m2 and chronic hunger is proxied by BMI <18.5 kg/m2. ‘n/a’ = not applicable, ‘–’ = not relevant to the scenario.
Source: Vivid Economics.
