EQUITY IMPACTS OF PRICE POLICIES TO PROMOTE HEALTHY BEHAVIOURS

Abstract

Governments can use fiscal policies or regulation to influence the prices of products with potential health impacts, aiming to change their consumption. In particular, policies aimed at raising prices, such as taxation, have caused concerns because they may impose an unfair financial burden on low-income households. We estimated patterns of expenditure on potentially unhealthy products by socioeconomic status, based on household expenditure surveys, with a primary focus on low- and middle-income countries, and we found that price policies affect the consumption and expenditure of a larger number of high-income than low-income households, and any resulting price increases are financed disproportionately by high-income households. As a share of all household consumption, however, price increases are often a larger burden for low-income households, depending on how much consumption is changed, most consistently in the case of tobacco. Larger health benefits will likely accrue to individual low-income consumers, due to their stronger response to price changes, but depending on initial consumption and associated health risks. In the case of taxing unhealthy products, a potentially larger financial burden on low-income households can be mitigated by a pro-poor use of the tax revenues generated.

What are price policies for health promotion and is their use justified?

The framework for addressing a rising global epidemic of non-communicable diseases (NCDs) adopted at the 2011 High-level meeting of the UN General Assembly ¹ identifies four leading behavioural risk factors for NCDs: tobacco use, harmful alcohol consumption, poor diet and insufficient physical activity. At least three of these risk factors involve the consumption of products purchased in markets that national governments regulate to varying degrees. Through regulation and fiscal policies (both taxes and subsidies), governments can affect market prices, turning them into behavioural incentives for health improvement. In the pursuit of the Sustainable Development Goals (SDGs) targets on NCD mortality, greater attention is turning to policies that can help achieve that target, and to the means required to finance NCD prevention and control. Price policies, and particularly, higher taxes on tobacco and other potentially unhealthy products may support both.

The benefits of addressing tobacco and harmful alcohol use, and improving diet include reductions in the major NCD killers, such as cardiovascular disease, stroke, cancer, and chronic respiratory diseases, as well as injuries and alcohol-use disorders. In most countries, these conditions affect the poor disproportionately.^1a Population-level interventions to address those risks have been found to be more cost-effective than medical interventions to treat diseases.^2–5

Taxes on tobacco and alcohol are key components of the Framework Convention on Tobacco Control and of WHO’s Global Strategy on Tackling Harmful Alcohol Use.^11–12 Taxes on salt have been used for centuries, but generally not for public health purposes. In a number of countries, taxes on food and non-alcoholic beverages have been adopted at the national or local level as part of efforts to improve nutrition and prevent obesity,^13–14 and calls are being made to extend the use of taxes in this area. Higher fast-food prices have been found to be associated with lower body weight, especially in young consumers; and reduced fruit and vegetable prices are associated with lower weight in low-income groups.¹⁵

In addition to using fiscal policies, many countries have also tried to regulate the minimum price of alcohol and tobacco products, as well as the use of price promotions to increase sales. A 10% increase in the minimum price of alcohol in Canada has been estimated to reduce acute alcohol-attributable hospital admissions and chronic alcohol-attributable admissions 2 years later, both by about 9%.¹⁶

Taxes are also an important source of revenue. Tobacco taxes were described in the Addis Ababa Financing for Development outcome document as offering “…a revenue stream for financing for development in many countries.”^16a In 2015, tobacco and alcohol excises contributed to around 1.2–1.7% of total tax revenues in Argentina, Denmark and Belgium, around 3.3% in the United Kingdom and Hungary, and more than 9.5% in Venezuela (OECD.stat). A tax on foods rich in saturated fat collected in Denmark between 2011 and 2012 accounted for 0.14% of total tax revenues. A recent WHO estimate showed that raising cigarette taxes by 50% in low-income countries like Congo, Lao People’s Democratic Republic, Viet Nam, or Madagascar would generate additional revenues equivalent to over 25% of current government expenditure on health. Raising taxes on alcohol to 40% of beverage retail prices would have similar, or bigger, effects .¹⁷

Price policies, in theory, interfere with the functioning of markets and with individual choice, but they can be justified when markets do not operate efficiently. This is typically the case, for instance, when consumers do not bear the full cost of their choices because the market prices of the products they consume do not reflect harms to others (e.g. from second-hand smoking, alcohol-related traffic accidents or violence) or the extra health care costs borne collectively,^18–20 so the consumption of those products tends to be higher than socially desirable. The addictive properties of some products and the influence of commercial advertising contribute to further increasing consumption.

Evidence that consumers respond to price incentives justifies the expectation that price policies will generate beneficial health impacts. However, the equity impacts of price policies have been a concern for many governments, because of the risk that the poor may be affected disproportionately. There is a widespread perception that expenditures due to increased prices weigh more heavily on the incomes of people of lower socioeconomic status, although the same people may also benefit more than others in health terms. There is also some evidence that higher tax revenues (including those collected through goods and sales taxes) are associated with higher public health expenditures, and consequently with better health outcomes (Carter and Cobham 2016 Are taxes good for your health? WIDER Working Paper 2016/171).

This paper assesses the existing evidence and provides new results based on household survey data sources on the equity impacts of price policies on potentially unhealthy products, in terms of both their financial and health consequences on individuals and households in countries at different levels of income. The final section of the paper brings together different sets of findings to help policy makers in different countries decide whether concerns about the equity impacts of price policies are a legitimate barrier to the use of these policies in the pursuit of health goals. Many examples in this paper are based on taxes, as these are most commonly used, but our analyses and conclusions are applicable to a broader range of price policies.

Socioeconomic patterns of potentially unhealthy consumption

We start by considering how consumption differs between socioeconomic groups in order to gauge the likely impacts of price policies. We have assessed patterns of consumption by socioeconomic status for four aggregates of products: tobacco products (hereinafter tobacco); alcoholic beverages (alcohol); non-alcoholic beverages, excluding water (soft drinks); and, snacks and confectionery products (snacks). These aggregates differ slightly between countries, in terms of the products they contain. A complete set of definitions is provided in the Web Annex. Socioeconomic status is defined as a function of income or total (expenditure-based) household consumption, depending on data availability (hereinafter Income, for ease of reference). Results presented in the main body of the paper are primarily in the form of top-to-bottom quintile ratios, i.e. as ratios between the average value of the relevant outcome in the top quintile and the average value in the bottom quintile. Results for individual quintiles are fully reported in the Web Annex. These show that in a small number of countries the relevant outcomes are not uniformly increasing or decreasing across socioeconomic groups, which should be considered when interpreting top-to-bottom quintile ratios.

The countries covered in the analysis were selected on the basis of the availability of good quality data from relatively recent expenditure surveys (undertaken in or after 2000). These countries provide wide geographical coverage across the following regions: Latin America (Chile, Guatemala, Panama, Nicaragua); Central-Eastern Europe (Albania, Poland, Turkey); Central Asia (Tajikistan); Sub-Saharan African (Tanzania); West Africa (Niger, Nigeria); and South and East Asia (India, Timor-Leste). These countries cover a wide range of national income levels, from high (Chile and Poland) to low (Niger, Tanzania), but exclude the highest-income OECD countries.

Prevalence of consumption and average expenditure on potentially unhealthy products in different socioeconomic groups

Top-to-bottom quintile ratios of consumption prevalence rates for each of the four aggregates of products is shown in Figure 1. Prevalence rates include all households with a positive (greater than zero) expenditure over the period covered by each survey, based on data from four national surveys (Chile, India, Poland and Turkey) and from a subset of the Living Standards Measurement Surveys (LSMS), an international collection of harmonised national household expenditure surveys supported by the World Bank. A parallel set of bars in the same figure shows top-to-bottom quintile ratios of expenditures for the four groups of products, only for consumers (i.e. excluding households that do not consume those products) from the same surveys. Differences in expenditure levels between groups may reflect not only different levels of consumption, but also different price levels paid by different groups of consumers (e.g. because they consume a more or less expensive mix of products within the relevant aggregate). The latter differences are likely to be small in relatively homogeneous product aggregates (e.g. soft drinks), but larger in more heterogeneous aggregates (e.g. alcohol).

Figure 1.

Socioeconomic disparities in prevalence of consumption and household expenditure on four product categories.

Notes:

1. Ratio of household expenditure on the specified product category in high SES group(top wealth quintile) over expenditure in low SES group (bottom quintile) among consumers only.

2. Ratio of prevalence of consumption in high SES group (top wealth quintile) over prevalence in low SES group (bottom quintile).

Ratios are calculated after logarithmic transformation of prevalence and expenditure values. Ratios above zero indicate larger prevalence or expenditure in the top wealth quintile; negative ratios indicate larger prevalence or expenditure in the bottom quintile.

Tobacco

There is no clear or consistent pattern of prevalence of tobacco use by socioeconomic status in the selected countries (Figure 1a). Guatemala, among the countries examined, displays the clearest gradient, with people in the top consumption quintile 2.4 times as likely to use tobacco products as those in the bottom quintile. Individual behaviours cannot be gauged from household expenditure data, but other data sources can shed light on patterns of tobacco use for men and women. For instance, data from the Demographic and Health Surveys (DHS) relying on individual self-reports, show a clear inverse relationship between socioeconomic status and tobacco use in men in several countries, and an opposite pattern in women (including in Albania, where no clear gradient is observed in household-based prevalence data)²¹.

In contrast, top-to-bottom quintile ratios of tobacco expenditure (also in Figure 1a) show consistent gradients across countries, with significantly larger expenditures by wealthier households. Overall, the information provided in Figure 1a for the countries examined suggests that policies aimed at altering the prices of tobacco products would not systematically affect a larger, or smaller, number of poor households. In absolute terms, a larger burden will be borne by wealthier households, given their larger expenditures on tobacco products (over six times as large as that of low-income households in countries like India, Timor-Leste and Guatemala).

Alcohol

Figure 1b, on alcohol use, shows much clearer socioeconomic patterns. Higher socioeconomic status is associated with a higher prevalence of alcohol use in most of the countries covered in our analyses (with India and Tanzania displaying a gradient in the opposite direction and Nicaragua showing a mixed gradient). Data from household expenditure surveys do not necessarily reflect the patterns of drinking that put people most at risk, such as binge drinking and regular heavy drinking. Other data sources indicate that harmful drinking is most common among men of low socioeconomic status, and women of high socioeconomic status, at least in many OECD countries.²² The gradient in average expenditure on alcohol is significantly steeper, with wealthier households spending larger amounts than households at the bottom of the socioeconomic scale (over four times as large in Timor Leste and Tanzania, and over seven times as large in India). Based on these findings, policies aimed at increasing the prices of alcoholic beverages will affect fewer poor, than rich, households, and their burden, in absolute terms, will fall disproportionately on the better off, although heterogeneity in patterns of consumption means that some poor households may still bear a large financial burden.

Soft drinks and snacks

Although data are available for fewer countries, soft drink use by households appears to follow similar patterns to those observed for alcohol (Figure 1c). Both the prevalence of soft drink use and expenditures on those beverages are greater in wealthier households, with somewhat less steep gradients in household expenditure than those observed for alcohol, but with up to six-fold differences between socioeconomic groups in India and Niger. It must be noted, however, that the soft drinks aggregate is broad and includes different types of beverages, e.g. with added sugar, with naturally occurring sugars, and with artificial sweeteners – potentially with different patterns of use by socioeconomic group – and some of these may not be the targets of price policies. Patterns are less clear for snacks (Figure 1d), but expenditure gradients are consistent with those observed for soft drinks, with larger expenditures by wealthier households, and India again displaying the largest difference (over six-fold).

Figure 1 shows that patterns by socioeconomic group vary across countries, but without a clear correlation with income or geography. Other country characteristics, presumably associated with national culture and traditions, seem to play a more important role.

Distribution of the health outcomes of price policies

The data examined so far provide evidence of the shares of households whose consumption would be affected by price policies. However, the health impacts of price policies are determined, above all, by the degree to which consumers respond to price changes (“price elasticity of demand”, in economic terms), and by the substitutions that consumers may make following changes in the prices of the products they purchase. Evidence of consumer responses to price changes is not available in all countries and for all products. In this section we present a summary of relevant existing studies.

For most of the products targeted by taxation or price regulation, the proportionate change in consumption to be expected is generally less than the proportionate change in price caused by the policy (“inelastic” demand). For some products – e.g. sugar-sweetened beverages¹⁵ – and some population groups – e.g. young smokers – a proportionately larger change in consumption has been observed. An inelastic demand, which is partly due to addiction, at least in the case of tobacco and alcohol products, tends to be associated with more limited substitutions by consumers; a higher likelihood of the tax being passed on to consumers (as opposed to being absorbed by suppliers); and larger revenues for government. On the other hand, an inelastic demand also means that a price intervention has to be relatively large to elicit a response that may lead to meaningful health gains.

Most importantly for the purposes of this paper, the size of consumer responses to price changes is a strong indicator of potential health gains, and different responses by people in different socioeconomic groups mean that a price policy may have effects on the distribution of health gains across socioeconomic groups. For instance, a greater response in lower socioeconomic groups, which is often observed due to tighter budget constraints, is also an indicator of pro-poor health outcomes, although not the only one, because health outcomes will also be a function of initial consumption and the level of risk associated with it.

Tobacco

The demand for tobacco products in low- and middle-income countries is at least as responsive, and often more responsive, to price than it is in high income countries,²³ with some studies suggesting it could be twice as responsive in low- and middle-income countries.²⁴

Within countries, there is evidence of a greater response to price changes by young and low-income consumers, although evidence of a socioeconomic gradient in price elasticity is less consistent in low- and middle-income countries.^23,24 In China, for instance, estimates suggest the response is five times as large in the bottom income quintile as that in the top quintile, and twice as large in people below age 24 as in those above age 65.²⁵ In Bangladesh, based on a three-level income classification, low-income cigarette smokers respond to price changes at least twice as strongly as high-income smokers .²⁶ In both countries, the socioeconomic gradient in consumer responses to price changes is largely due to different changes in smoking participation, with virtually no gradient in demand reductions by those who continue to smoke .^26,27

Alcohol

A consistent body of evidence shows that increases in the prices of alcoholic beverages reduce alcohol consumption.^28–31 Alcohol drinkers’ responses to price changes were found to be similar in countries at all levels of income.³² Despite widespread belief that the response to price changes of low-income drinkers is stronger than that of high-income drinkers, and at least some evidence of this,³³ empirical assessments of the size of a possible socioeconomic gradient in response, and how the gradient may vary across countries at different levels of income and development, remain limited .³⁴ The response to the possible introduction of a minimum price for alcohol in the United Kingdom was estimated to be 7.6 times as large by drinkers in the lowest income quintile as by those in the highest.³⁵

Food and non-alcoholic beverages

A large number of systematic or structured reviews have recently synthesised the evidence on the effect on consumption or sales of relevant food prices.^16,36–42 Few comprehensive, reviews include evidence from LMICs. Green et al.,⁴¹ focuses on consumer responses as changes in their consumption of the products whose prices have changed (“own-price elasticity”), based on evidence from 136 studies and 3,495 estimates, while Cornelsen et al.⁴² examines the much less frequently assessed substitutions that may occur as a result of price changes (“cross-price elasticities”), based on 78 studies and 4,162 estimates. Nakhimovsky et al. focuses on own-price elasticity of sugar-sweetened beverages in MIC.

Increases in the prices of food and non-alcoholic beverages elicit the greatest changes in consumption in low-income countries, followed by middle- and then high-income countries. Within countries, price elasticities are higher in the lowest-income than in the highest-income groups, but differences are relatively small, with a response by lowest-income consumers between 1.14 and 1.21 times as large as that of highest income consumers, for different food and beverage products. The evidence on within-country differences, however, is dominated by high-income-country studies, and it is limited to 21 studies.⁴¹ Results are similar for sugar-sweetened beverages, with one study estimating that the reduction in consumption among the consumers in the lower socio-economic third is as high as 9.1% while it is 5.5% among those in the highest third (Nakhimovsk et al. 2016, Taxes on Sugar-Sweetened Beverages to Reduce Overweight and Obesity in Middle-Income Countries: A Systematic Review).Panel 2 provides an illustration of the greater response of low-income consumers, based on an evaluation of a tax on sugar-sweetened beverages recently introduced in Mexico.

Additional factors affecting the health impacts of price policies

The evidence presented in this section points to a larger reduction in consumption, which will likely lead to larger health gains, in people of low socioeconomic status than in people of high socioeconomic status, which partly depends on the respective initial levels of consumption and risk. This social gradient is especially steep for tobacco in high-income countries, and less consistent in low- and middle-income countries. Similar gradients also exist in consumer responses to changes in alcohol prices, although empirical evidence of these is more limited, and in responses to food and non-alcoholic beverage price changes, but in the latter case the gradients are significantly smaller. The size and type of substitutions that consumers will make is to a large extent the result of how price policies are designed (especially how the tax base is defined, i.e. what products are targeted) and evidence that substitution patterns differ between socioeconomic groups is very limited at present.

While health gains for consumers can be expected to be larger in low than in high socioeconomic groups, aggregate health gains may not have the same distribution. This is because of differences in the prevalence of consumption across socioeconomic groups, which is especially significant in the case of alcohol. However, more factors may contribute to determining aggregate health outcomes in different socioeconomic groups; examples include differences in patterns of consumption and the degrees of risk associated with them, access to care, and concurrent exposures (e.g. environmental factors). Therefore, the long-term outcomes of price policies are currently best estimated through mathematical models. A recent example is the Chronic Disease Prevention (CDP) model developed by OECD and WHO, which estimated that a package of fiscal policies including taxation of foods high in fat and subsidies on fruit and vegetables would lead to larger aggregate health gains in people of low socioeconomic status than in those of high socioeconomic status.^4,43

Financial impacts of price policies in different socioeconomic groups

Governments’ concerns about the potentially regressive financial effects of price policies for health promotion have been among the main barriers to a wider use of such policies. However, the meaning of “regressive” is not always clearly or consistently defined in the public debate, beyond its generic association with a larger financial burden placed on low-income than on high-income individuals or households. In the case of taxes, for instance, the most common measure of financial burden in a given socioeconomic group is the average ratio between taxes paid and ability to pay (generally income or disposable income, but also total household expenditure), across all households in that group. This measure does not take into account the positive financial effects of taxes through the health improvements they may generate. Also, this tax burden measure is heavily influenced by the proportion of households that consume the taxed product in each group, showing a lower tax burden in groups in which fewer households consume the product. Reducing the prevalence of consumption is a goal of price policies for health promotion, so it is important to account for prevalence reduction as a factor that helps to contain the burden of taxation (or price policies more generally) in a given socioeconomic group. The main limitation of a burden measure encompassing all households is that it does not reflect the real burden borne by households whose members consume the product. To determine this, households that do not consume the product in question should be excluded from group averages, which we have done in the calculation of a second measure of the financial burden of price policies (“consumers only”), shown alongside the first. We rely again on household expenditure survey data from the countries previously examined.

A further important distinction is between the distribution of the burden of a price policy and the distribution of the burden caused by a change in an existing policy. Tobacco and alcohol taxes, for instance, are almost universal, and the relevant policy question today is not whether these products should be taxed, but whether existing taxes should be increased. A tax can be regressive, yet an increase in the same tax might attenuate its regressive impact if the response by low-income households were sufficiently larger than that of high-income households. A weak response (typical of high-income groups) is associated with a larger increase in expenditure on a taxed product (and therefore a larger increase in tax paid). This is illustrated in Panel 1, using the example of a tobacco tax hike in Lebanon. Of course, these are average effects, summarising individual situations ranging from a drop in tax burden for those who stop consuming the taxed product, to an increased burden for those who are least responsive to price changes.

As an extension of recent OECD work on the impacts of consumption taxes in different socioeconomic groups⁴⁴ we present in Figure 2 estimates of the distribution of the burden of excise taxes on alcohol and tobacco in three OECD countries: Chile, Poland and Turkey (selected as the lowest-income countries in the OECD study, based on GDP per capita, for which data were available; two of these, Poland and Turkey, are also part of a group of 23 high NCD burden countries identified in previous Lancet series^44a). For these countries, tax burden is measured as a ratio between tax expenditure and total household expenditure. However, the numerator of this measure is not available for other low- and middle-income countries examined in this paper. For the latter, we present distributions of the proportion of household expenditure on tobacco, alcohol, soft drinks and snacks, which is a proxy of tax burden.

Figure 2.

Tax burden of tobacco and alcohol excises in 3 OECD countries.

Tobacco

Figure 2a shows the distribution of the two measures of tax burden (all households and consumers only), for tobacco, by income quintile. In the three countries in Figure 2a, the two approaches lead to similar conclusions regarding the distribution of the burden of tobacco taxation, because differences in the prevalence of consumption between socioeconomic groups are very small (prevalence increases slightly with income in Chile and decreases slightly in Poland and Turkey). In all three countries, households in the bottom income quintile bear roughly twice as large a burden from tobacco taxes as that borne by households in the top quintile.

Alcohol

The picture for alcohol (Figure 2b) is very different, because of the significantly larger prevalence of alcohol use in higher than in lower socioeconomic groups. The burden of alcohol excise taxes averaged across all households in each quantile is progressive in all three countries, with households in the top income quintile bearing a burden between 1.6 (Chile) and 2.8 (Turkey) times as large as that borne by households in the bottom quintile. However, the tax burden borne by households that do consume alcohol is very slightly regressive in Chile and Poland and more steeply regressive in Turkey, with 2.4 times as large a burden in the bottom quintile as in the top quintile. This means that the burden of alcohol taxes is borne disproportionately by higher-income groups, but the financial burden borne by individual low-income households consuming alcohol is proportionately larger than that borne by high-income households consuming alcohol. This is consistent with the findings of a UK study, which showed a progressive pattern for a 5% increase in alcohol prices, in terms of average impacts across all households, and a regressive pattern when the denominator included only households consuming alcohol .⁴⁵

Figure 3 shows the weight of each of the four product categories on total household expenditure in different socioeconomic groups, in the same countries shown in Figure 1 ( averages across all households and consumer households, respectively). Tobacco products (Figure 3a) account for a larger proportion of low-income than high-income households’ expenditure, with few exceptions (Tajikistan, Timor-Leste, Panama). Averages across all households and consumers only are similar in most cases, because gradients in the prevalence of tobacco consumption are not steep or consistent in the countries concerned. Conversely, alcoholic beverages account for a larger share of the expenditure of high-income households, as an average across all households in each socioeconomic group, in all countries except Tanzania, Nicaragua and Guatemala. However, alcoholic beverages weigh disproportionately on the expenditure of low-income households that consume alcohol in most countries.

Figure 3.

[7–10 in draft paper]: Top-to-bottom income or wealth decile ratio (ln) of expenditure on product category as a share of total household expenditure.

Soft drinks and snacks

Average expenditures on soft drinks across all households in each income quintile (Figure 3c), as a share of total expenditure, provides an indication that price policies would have progressive or neutral impacts in most countries, with Guatemala and Nicaragua displaying regressive patterns. The prevalence of consumption is lower in lower-income quintiles, and the financial burden on consumer households consistently shows a less progressive, or even regressive, distribution. In the case of snacks, a regressive distribution of expenditure shares averaged across all households is seen in Guatemala, Nicaragua, Panama and Niger (Figure 3d), while the distribution is progressive in the Asian countries in the figure and in Tanzania. However, in at least some of the countries, the income groups bearing the highest potential burden of price policies are those in the central part of the distribution. Because the prevalence of consumption increases with income, the burden on consumer households is more regressive, with a progressive distribution being preserved only in Tajikistan.

In summary, the burden of price increases will potentially be proportionately larger in lower socioeconomic groups in the case of tobacco, and in higher socioeconomic groups in the case of alcohol, soft drinks and snacks, albeit with exceptions. However, when considering only households that consume the above products, the burden tends to be higher in lower socioeconomic groups.

Should governments be concerned about the financial impacts of price policies?

The use of price policies by governments for the purpose of improving health has generated a large debate, focused on both the benefits and the unintended consequences of such policies. The potential for price policies to have regressive financial impacts has been one of the main arguments against the use of price policies, often used as part of opposition efforts by industry stakeholders. This paper is an attempt to scrutinise existing data that may help governments to understand the equity impacts of possible price policies on tobacco, alcohol, soft drinks, and snacks. The effects of price policies are not assessed directly in our analyses, because the data that would be required are not available in the countries concerned. Rather, we have relied on household expenditure data and information on consumer responsiveness to price changes to gauge the likely distribution of the impacts of price policies across socioeconomic groups.

While it is not possible to draw a simple and generalizable conclusion from the analyses presented here, a number of important findings focused on a selection of predominantly low- and middle-income countries can provide helpful guidance to governments. On the whole, these findings suggest that concerns about adverse equity impacts may be outweighed, in many cases, by the expectation of health gains and by a number of pro-poor effects.

More high- than low-income households are affected by price policies

The first finding is that price policies will affect a larger number of high-income than low-income households, and the absolute increases in expenditure involved will be larger for high-income households. This is because the prevalence of consumption and the expenditure on alcohol, soft drinks and snacks increase consistently with household income. This is also true for expenditure on tobacco products, although there is no clear and consistent gradient in the prevalence of tobacco use in the countries examined. An important implication of this conclusion is that the extra expenditures caused by policies that increase prices (e.g. the revenues generated by taxes) come disproportionately from high-income households. Tax design can influence this effect, and tax policy makers have to consider potential tradeoffs. For instance, a volumetric tax on alcohol may be more effective in reducing the number of units of alcohol consumed, but an ad valorem tax (a function of price) would likely shift more of the financial burden onto higher-income consumers.

Low-income households often (not always) bear a larger tax burden

The second finding concerns the distribution of the burden of tax policies as a proportion of total household expenditure. Low-income households bear a larger tobacco tax burden, consistently across countries. The distribution of the tax burden of alcohol taxes is generally progressive, although the burden borne by just the low-income households that consume alcohol is proportionately larger than that faced by high-income households consuming alcohol. The potential distribution of the burden of price policies targeting soft drinks and snacks varies between countries, but again, the low-income households consuming these products tend to bear a larger financial burden.

Low-income consumers enjoy larger health benefits

Finally, the health benefits generated by price policies will be larger for those affected low-income consumers, due to their stronger response to price changes, but governments have to design their policies carefully in order to minimise unhealthy substitutions in consumption. At the aggregate level, larger gains may still accrue to higher-income groups where there is a higher prevalence of consumption in these groups.

The data for this paper, mostly based on household expenditure surveys, are readily available; nevertheless, they have a number of limitations, typically resulting from the self-reported nature of expenditure information and from product aggregations. In addition, the household-based nature of these datasets means that we are not able to identify individual-level patterns, which are important when assessing the potential health impacts of price policies, and also to determine the prevalence of consumption associated with higher levels of risk. The availability and reliability of household expenditure data vary greatly across countries and tend to be better in higher-income countries, but data on consumer responsiveness to price changes and likely substitutions in different socioeconomic groups are even scarcer.

Health Impacts have financial consequences too

Health impacts, in turn, may have financial consequences for the households concerned. For instance, the analysis in Panel 1 shows that tobacco taxation can reduce the incidence of catastrophic health care expenditures in low-income households. Studies undertaken by the World Bank on tobacco taxation in countries such as Armenia^45b and Chile^45c have shown that taxes can increase, rather than decrease, household net incomes through lower out-of-pocket medical expenses and higher earnings due to an increase in working years, in addition to preventing catastrophic health expenditures and poverty in low-income populations.

Whatever the measure of equity impact, price policies are never unequivocally regressive

We have shown that the direction and size of the equity impacts of price policies depend to a large extent on what measures are used to assess them and on the specific objectives of government policy. In no case are these impacts unequivocally regressive. In addition, where taxes or tax increases do generate regressive tax burdens, governments should also consider the positive financial impacts linked with health improvements triggered by taxation, as well as the size of the monetary impacts involved. For instance, possible taxes on sugar-sweetened beverages in the United States and in Australia were estimated to weigh more heavily on the poor, with relatively steep gradients in terms of share of tax burden on household income, however, the extra tax payments were estimated to be 4 USD per low-income household per year for a 0.5 USD/cent per ounce tax in the United States, and up to 3.7 AUD for a 0.2 AUD per litre tax in Australia (averages across all households in each income group).⁴⁶⁻⁴⁷ Of course no amount of money, however small, is trivial for low-income households, especially in low-income countries, but governments must consider carefully whether the amounts involved represent a barrier to the implementation of a potentially beneficial health policy, from which lower-income people are likely to benefit disproportionately.

A further reason why regressive impacts such as those described in this paper and elsewhere require a cautious interpretation is that at least the most common type of price policy, i.e. taxation, generates significant amounts of government revenues, which can be used to mitigate or even offset any unfair distribution of tax burden. Whether or not tax revenues are earmarked – and there are several important examples of earmarking, two of which (Thailand and Philippines) are illustrated in the Web Annex – the revenues generated will contribute to the larger tax revenue pool, typically used by governments to deliver public services, of which low-income people may benefit disproportionately.

In China and in India, it has been shown that a 50% increase in tobacco prices would lead to larger decreases in expenditure on tobacco-related diseases, as a share of income, in the bottom income quintiles than in the top ones, providing financial risk protection to those who have the lowest incomes.^25,48 The paper by Jan et al. in this series provides further examples of how revenues can be used by governments for risk protection .⁴⁹ Finally, what governments are, or should be, mostly concerned about is the regressive or progressive nature of their tax system as a whole, not individual tax measures. When Denmark introduced an innovative tax on saturated fat in 2011, it did this as part of a broader tax reform that altered the progressivity of its income tax. ⁵⁰

Our analysis provides some clear guidance on a number of impacts that countries can expect from price policies on the four groups of products examined. However, a final lesson from the analyses is that a detailed assessment is required in each setting and for each policy, because important dimensions of impact cannot be generalised and uniform patterns cannot always be identified. The social dimensions of consumer behaviours vary between and within countries. Claims that are not based on a detailed analysis of country-specific data are bound to be simplistic and most likely misguided.

Key Messages.

1. Price policies on potentially unhealthy products alter consumption and expenditure for all consumers. Health and economic impacts differs by SES groups.

2. Expenditure on potentially unhealthy products increases more for high-income households than low-income households in response to a price increase. But this change in expenditures is often a heavier burden for low-income households because it constitutes a higher share of their overall expenditure. This is most consistent in the case of tobacco products.

3. Larger health benefits are likely to accrue to individual low-income consumers, due to their stronger response to price changes.

4. What is meant by regressive impacts are not always clearly or consistently defined in the public debate. Different measures may lead to different policy conclusions (most notably in the case of alcohol) and policy makers must be aware of such differences.

5. Adverse equity impacts of taxes can be mitigated by a pro-poor use of the tax revenues generated, or by adjustments in the distributional effects of the broader tax system.