How prices and income influence global patterns in saturated fat intake by age, sex and world region: a cross-sectional analysis of 160 countries

Data and sources

We used secondary data sources for the analysis. SF intake data measured in per cent of total energy per day (% energy/d) for a representative individual was obtained from the 2018 GDD. The GDD, maintained by the Global Nutrition and Policy Consortium at Tufts Friedman School of Nutrition Science and Policy, provides comprehensive and comparable dietary intakes for major foods and nutrients in 185 countries and territories. The GDD was developed using systematic searches of available survey data on individual-based dietary intakes for key food and nutrient categories at the national and subnational levels. GDD intake estimates are based on the results of existing surveys (1248 in total), representing 188 countries and approximately 99% of the global population. It is the first database to provide estimates of daily consumption levels by food or nutrient category and contains representative individual intake data by age (0–1 year, 1–2 years, 2–5 years and then by increments of 5 years to age 97.5) and sex.26 The GDD also disaggregates individual intakes by three education levels and residence (urban and rural). The GDD data estimation process included extensive communication with researchers and government authorities and large subnational surveys when other options were unavailable.27 28 For details on the GDD coverage, data methodology and data collection, see https://www.globaldietarydatabase.org/methods/summary-methods-and-data-collection.

National food expenditure and price data from the World Bank International Comparison Programme (ICP) were used to derive an SF price series. Although our intake measure is comprehensive and inclusive of all food sources, the price series used for the analysis was limited to the primary contributing food categories: meats, dairy, and oils and fats. The price series for the meats category in the ICP database is an aggregation of the following: beef and veal; pork; lamb, mutton, and goat; poultry and other meats and meat preparations. Dairy—fresh milk, preserved milk and other milk products, cheese and curd, and eggs and egg-based products. Oils and fats—butter and margarine and other edible oils and fats.29 Although SF is readily found in a wide array of foods, these categories have been identified as major contributors to saturated fatty acids in diets.30 While other foods, such as sweet and savoury snacks, also contribute and are included in our SF intake variable, global price series for these food categories are not widely available.

The ICP is a global initiative that estimates purchasing power parities (PPPs) and price level indices (PLIs) across countries, which allows for global comparisons of spending and economic well-being. PPPs are spatial price deflators that make it possible to compare expenditures across economies.31 PLIs are PPPs standardised to a common currency (generally the US dollar) or indexed to a global average or base country.32 The most recent ICP data round (2017) included comparative prices and expenditure data from 176 participating economies.32

For income, we used 2018 PPP-adjusted, gross domestic product (GDP) per capita from the World Development Indicators (WDI) database. Because differences in currency values and exchange rates do not always consistently reflect price-level differences across countries, PPP-adjusted GDP allows for cross-country comparisons because overall price disparities across countries are taken into account.33

The analysis was limited to the 160 countries represented in all three databases (GDD, ICP and WDI), which are listed in online supplemental table 1 by geographical region (see the online supplemental file 1): East Asia, Southeast Asia, and Asian Pacific (Asia) (14 countries); Central and Eastern Europe and Central Asia (CEE) (27 countries); Latin America and Caribbean (LAC) (29 countries); Middle East and North Africa (MENA) (17 countries); South Asia (S-ASIA) (seven countries); Sub-Saharan Africa (SSA) (43 countries) and high-income/rest of world (HIC) (24 countries). HIC is an aggregation of HIC in the Western hemisphere, Australia and New Zealand, with the addition of a few surrounding islands. Countries without data in any of the three databases were excluded.

Supplemental material

See the online supplemental file 1 for a more detailed discussion of the price, expenditure and income data by geographical region.

Patient and public involvement

We used secondary data for this study. All data are publicly available and did not require direct patient involvement in the study design or implementation.

Model and estimation

To estimate SF intake demand, we used a semilog functional form that has been proven to be consistent with economic theory and rational consumer behaviour.34 35 Many studies have used a double-log form.36 However, a problem with the double-log form is that significant intake differences across subgroups can be lost in log conversions. A semilog relationship allowed for a better assessment of subgroup effects on intake responsiveness. Also, it has been shown that semilog models contain the necessary information for obtaining, for instance, reliable measures of consumer welfare and the underlying preference structure of consumers.34 Prior studies have also used a demand-system approach, primarily due to the adding-up property when using expenditure data (ie, expenditures on all food categories ‘add up’ to total food expenditures), which results in the error terms being correlated across equations specific to each food category. Since this relationship does not exist with individual intakes, particularly when the correspondence between purchases and intakes is not one-to-one, we can estimate intake demand for a single food or nutrient category separately.19 20

Let qgC represent the % energy/d from SF for demographic subgroup g (g: sex and age) in country C, and let pC represent the price level index for the contributing food categories in country C. Let YC and PC represent real per capita income and the food price level index, respectively, in country C. Given these terms, the following model was used to estimate the relationship between intake, income and prices:

Embedded Image

(1)

The Embedded Image

terms (k={0,1,2}) are parameters to be estimated, and ugC is a random error term. Note that the price term is defined by the price of contributing food categories (pC) relative to overall food prices (PC). Thus, the model discounts any price differences across countries due to differences in overall food prices and implicitly accounts for the cross-price effects of other foods. For instance, if dairy prices were the same in two countries but overall food prices differed, intake would be greater in the country with the higher food price level since dairy is relatively cheaper when compared with food overall. Note that equation (1) does not include higher-order income and price effects (eg, quadratic income and price–income interactions). In preliminary analysis, these higher-order terms were highly insignificant, which implied that price or income responsiveness did not depend on the level of per-capita income.

Using equation (1), we estimated intake demand using a procedure that allowed for error correlations among observations from the same country (ie, country-clustered errors).37 To account for differences in preferences across countries due to cultural differences or other related factors, we included regional binary variables in the analysis (ASIA, CEE, LAC, MENA, S-ASIA and SSA). We accounted for age and sex by allowing these factors to have a direct effect on intake as well as an additional effect through income and prices. Thus, the beta terms Embedded Image

were expanded to account for age and sex interactions: Embedded Image. Further disaggregations (education level and residence) were not considered due to estimation concerns resulting from negligible differences in SF intake across these factors. Although we used a single price index (pC) to represent the three food categories (meats, dairy, and oils and fats), intake responsiveness with respect to the price of each food category was easily derived. Defining the conditional expenditure share and price for the ith food category in country C as siC and piC, respectively, pC is as follows: Embedded Image. Thus, the relationships between qgC and piC were derived using the estimate of the price term in equation (1) Embedded Image and the conditional expenditure share siC as follows: Embedded Image.

This post was originally published on https://bmjopen.bmj.com