Do tax waivers reduce prices and increase imports of healthy foods? Interrupted time-series analyses from Tonga

Abstract

Introduction

It is unclear whether tax waivers are a useful public health intervention for lowering food prices and improving nutrition, or whether fiscal gains go largely to importers, wholesalers and retailers. Tonga reduced and removed 10% to 20% import tariffs and consumption taxes on a selection of healthy foods in 2013, 2015 and 2016. This study investigates the impact of these changes on retail prices and import volumes of targeted foods.

Methods

Interrupted time series analysis was used to quantify the percentage changes in retail prices and import volumes in the first year after each tax waiver compared with a counterfactual based on pre-existing trends. A segmented linear regression model was run with adjustment for autocorrelation and the effects of gross domestic product, international visitors, month, exchange rate and international oil prices. Meta-analysis technique was used to summarise findings across foods.

Results

On average, there was a suggestion of a small decline in retail price of tax waivered foods (−4%, CI −10% to 2%) with an average pass-through to a reduced price for consumers of 51%. Two of nine relevant foods had statistically significant evidence of price declines: frozen vegetables (−20%, CI −25% to −14%) and cooking oil (−3%, CI −5% to 0%). Overall, import volumes of tax waivered foods did not increase, and instead, there was suggestive decline (−20%, CI −45% to 15%). No import changes were statistically significant.

Conclusion

While there was no evidence of increased food supply for tax waivered foods, significant price declines were observed for frozen vegetables and tinned mackerel. There was evidence of moderate pass-through of tax waivers to lower retail prices. Interpretability was limited by variability in estimates and a shift from imported to locally sourced foods. Further evaluation of tax waiver effects is needed.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Several countries have removed consumption taxes or import tariffs on selected healthy food items to improve health, but only a few have been evaluated for their effect on food prices, purchasing and consumption.

WHAT THIS STUDY ADDS

• After Tonga removed import tariffs on fish, vegetable oils, fresh and frozen vegetables and fresh fruit and consumption tax on fruit; the average decline in retail prices was small, with an average waiver pass-through to lower consumer prices of 51%. There was no increase in import volumes of waivered foods, which if anything suggested a non-significant decline.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Study findings do not provide any evidence for increases in the food supply of tax waivered foods; however, there was evidence of moderate average pass-through of tax waivers to retail prices and significant price declines were observed for two out of nine products. Further evaluation of tax waiver effects in other settings, including waivers with different policy designs, would help better understand the likely public health effects of tax waivers in practice.

Introduction

Fiscal policies, such as tax waivers on healthy foods, have been recommended by the WHO to promote healthy nutrition and prevent non-communicable diseases (NCDs).¹ Healthy nutrition, such as whole grains² and whole fruits and vegetables,³ is protective against cardiovascular disease, diabetes, cancer and premature death.

Several countries have removed consumption taxes or import tariffs on selected healthy food items to improve health, often targeting fruit and vegetables. But few tax waiver introductions appear to have been evaluated for their effect on food prices, purchasing and consumption. In one example from Fiji, import duty reduction on fruit and vegetables was followed by increased import volumes of duty waivered foods.⁴ This paper also suggests that the waiver may not have been passed through at supermarkets, and increased tourism may have driven study outcomes rather than local population consumption.⁴

Experimental and simulation studies examining consumption tax exemption on fruit and vegetables have had mixed findings. In an experimental study, the removal of consumption tax (20% value added tax (VAT)) on fruit and vegetables in a supermarket chain in Denmark led to 25% increases in revenue and 33% increases in the sales of fruit and vegetables in one supermarket chain, but the intervention also led to increased sales of other unhealthy and healthy foods (likely from attracting more customers).⁵ Modelling in Germany reports that the removal of 7% VAT on fruit and vegetables was found to increase demand by 5.2% for fruit, and 3.6% for vegetables, and reduce cardiovascular disease cases and deaths.⁶ An Australian fruit and vegetable subsidy ($0.14/100 g) was modelled to be cost-effective when added to a package of taxes, but did not necessarily lead to health benefits on its own, because of possible increases in consumption of foods high in fat and salt.⁷ In Norway, the effects of subsidising healthy foods were minor and were particularly limited among low-purchasing households.⁸

The literature evaluating the effects of subsidies in the form of discounts, vouchers and coupons,⁹ and rebates or cash back¹⁰ is more extensive. Systematic reviews and meta-analyses have not focused on tax waivers possibly because, as well as there being fewer studies, waivers are considered indirect measures and may not be wholly passed on to consumers.¹⁰ Meta-analyses evaluating more direct subsidies have reported significant increases in subsidised fruit and vegetable sales (waiver elasticity 0.59, CI 0.31 to 1.04) and a small but non-significant increase in consumption (waiver elasticity 0.17, CI −0.15 to 0.49).⁹ In another meta-analysis, price reductions of 20% were associated with increases in fruit and vegetable purchases by 17% (CI 12% to 21%).¹⁰ Another review observed that taxes on high fat, salt and sugar food may be more effective when combined with subsidies on healthy foods.¹¹

Arguments against the use of tax waivers have included high expected compliance costs¹² (eg, consumption tax exemption for fruit and vegetables when still being applied to other foods), limited pass-through to consumers¹³ and thus limited impact on consumption, potential increased consumption of unhealthy foods, and greater benefit for higher spenders who have the highest socioeconomic position.¹² Tonga’s tariff and consumption tax waivers enable the evaluation of many of these ideas in a real-world setting.

Tonga is a middle-income island nation in the South Pacific which has some of the highest obesity and diabetes rates in the world,¹⁴ a modern phenomenon driven by globalisation¹⁵ and the availability of imported food that is typically high in fat, salt and sugar. Tonga removed or reduced import tariffs on fish, vegetable oils, fresh and frozen vegetables and fresh fruit in 2013,¹⁶ 2015¹⁷ and 2016.¹⁸ In July 2016, fresh fruit was also exempted from the standard 15% consumption tax. Some tax waivers overlapped with the introduction of excise taxes on a range of foods high in fat, salt and sugar, especially in 2016.¹⁹ Nevertheless, public awareness about tax exemptions was very low (eg, 13% of consumers were aware of the 2016 fruit consumption tax exemption).²⁰

The aim of this study was to quantify the impact of tax waivers in Tonga on food prices and import supply after the waivers were introduced. Our focus was on waivered foods and potential substitute foods not subject to tax changes (eg, excise or waivers). If waivers are passed through to consumer prices, there would be decreased retail prices and expected increases in import volumes.

Methods

An interrupted time series analysis was used to compare food prices and import volumes in the 12 months after tax waivers were introduced compared with a ‘counterfactual’, which is the average expected projected outcomes if there was no tax change. The study analysis was pre-planned and published in the study protocol.²¹

Food tax waivers

Legislated tariff and consumption tax waivers in Tonga were introduced with health goals.²² In 2013, import tariffs were lowered on fresh, frozen and tinned fish (from 20% to 5%) and vegetable oils (20% to 10%)¹⁶; then import tariffs on fresh and frozen vegetables (20% to 0%) and fresh fruit (15% to 0%) were completely waivered in 2015,¹⁷ and the import tariff on potatoes and onions was completely waivered in 2016 (15% to 0%).¹⁸ In July 2016, Tonga also exempted fresh fruit from consumption tax (15% to 0%) (Personal communication Tonga Ministry of Revenue & Customs, 2024). These waivers were characterised by date, tax level and targeted foods (online supplemental table A).

The 2016 tax waivers overlapped with the introduction or increase of excise taxes on sugar-sweetened beverages,²³ turkey tails, mutton flaps, chicken leg quarters, instant noodles, mayonnaise and ice cream¹⁹ (ranging from 15% to 59% of import price). There was also some overlap of tariff waivers with earlier increases in taxes on sugar-sweetened beverages (27% increase)²³ and animal fat in 2013 and turkey tails in 2015 (the latter both 15% increases).

Datasets and variables

Retail surveys are used by the Tonga Department of Statistics to collect and monitor price data for commonly consumed foods. Prices of waivered foods were collected from three to nine retail outlets each month, and prices of locally sourced foods were also monitored weekly at local markets. Survey locations were either on the main island of Tongatapu or the Vava’u island group. Tonga Statistics supplied the study price datasets with monthly prices from September 2010 to August 2017 (online supplemental table B). Price trends for tax waivered foods and potential substitute foods were selected for analysis (online supplemental tables C and D).

Import volumes, values and tax revenue data were supplied by the Tonga Ministry of Revenue & Customs, for imported food items (HS 02-04,07-11,15-22) from January 2010 to December 2019. Each row accounted for a shipment of a particular food type categorised by the International Harmonised System (HS) of trade flow coding. Duplicate rows were removed. HS codes were then aggregated by month and food for waivered foods (online supplemental table A) and potential substitute foods (online supplemental table D). Data were checked to identify any unusual missing data, trends and outliers. An automated process was used to correct outlying data by recalculating volume from the warehouse description and doing manual checks of any remaining unusual values. This helped to ensure consistency of volume units (see study protocol for more detail).^{19 21}

Analysis

Generalised least squares (GLS) regression was used to fit a segmented linear model to monthly price and import volume data for each waivered food, with level and slope change variables for each tax waiver (see equation below: ‘β’ are the regression coefficients and ‘t’ is time). Observed post-waiver trends were compared with projected pre-waiver trends for 1 year. Testing and adjustment for autocorrelation and moving averages were informed by a plot of the autocorrelation function and partial autocorrelation function. If an autocorrelation adjusted model was significantly different from the unadjusted model (log-likelihood ratio test), then the model with the best fit to the data was selected (lowest Akaike Information Criterion (AIC)).

$$\begin{gathered} {\displaystyle {\displaystyle \begin{array}{rl}\mathrm{l}\mathrm{o}{\mathrm{g}}_{\mathrm{e}}\left(\mathrm{I}\mathrm{m}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{t}\mathrm{v}\mathrm{o}\mathrm{l}\mathrm{u}\mathrm{m}\mathrm{e} \\ \left[\mathrm{k}\mathrm{g}/\mathrm{p}\mathrm{o}\mathrm{p}\mathrm{u}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}/\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{h}\right]\right)& ={\beta }_0+{\beta }_1\times \mathrm{t}\mathrm{i}\mathrm{m}{\mathrm{e}}_{\mathrm{t}}+{\beta }_2\times \mathrm{t}\mathrm{a}\mathrm{x}\mathrm{\_}\mathrm{c}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}{\mathrm{e}}_{\mathrm{t}}\\ +{\beta }_3\times \mathrm{t}\mathrm{a}\mathrm{x}\mathrm{\_}\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{d}\mathrm{\_}\mathrm{c}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}{\mathrm{e}}_{\mathrm{t}}+{\beta }_4\times \mathrm{G}\mathrm{D}{\mathrm{P}}_{\mathrm{t}}\\ +{\beta }_5\times \mathrm{i}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{n}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{a}\mathrm{l}\mathrm{\_}\mathrm{v}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{\_}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}{\mathrm{s}}_{\mathrm{t}}+{\beta }_6\times \mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}{\mathrm{h}}_{\mathrm{t}}\\ +{\beta }_7\times \mathrm{e}\mathrm{x}\mathrm{c}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}\mathrm{\_}\mathrm{r}\mathrm{a}\mathrm{t}{\mathrm{e}}_{\mathrm{t}}+{\beta }_8\times \mathrm{o}\mathrm{i}\mathrm{l}\mathrm{\_}\mathrm{p}\mathrm{r}\mathrm{i}\mathrm{c}\mathrm{e}{\mathrm{s}}_{\mathrm{t}}+{\epsilon }_{\mathrm{t}}\end{array}}} \end{gathered}$$

The model was adjusted for potential (pre-selected) time-varying confounding by trends in gross domestic product (GDP), international visitor numbers, month (adjusts for season), T$/US$ exchange rate, international oil prices (proxy for shipping costs), and international food prices where available (for the oranges model only). Fresh fruit was subject to two sequential tax waivers, and these were evaluated in the same model with a level and slope variable for each waiver. Import volumes were log-transformed pre-modelling and post-modelling to avoid the prediction of ‘negative imports’ and stabilise the variance.

The observed outcome model was compared with the predicted counterfactual to calculate study outcomes. Multiplicative effects (rate ratios) were estimated and interpreted as a percentage change compared with counterfactual. Parametric multivariate simulation and bootstrapping were used to estimate 95% CI for the percentage change outcome (using the coefficients and their standard errors from the GLS model and allowing for correlation between correlates). Random effects meta-analysis technique was used to summarise outcomes across different excised taxed foods, with 95% CI. Sensitivity tests were run (post-hoc) to assess the effects of excluding each of the consumption tax waiver and the fresh and frozen fish waiver on the overall import volumes result. Further sensitivity analyses assessed the impact of model specifications on key study findings (online supplemental tables E and F). Suitable controls were not available for running pre-planned secondary analyses, but a pre-intervention placebo test was reported (online supplemental tables E and F).

The same analysis approach was taken to assess the impact of tax waivers on potential substitute or complementary foods, selected according to availability of both retail price and import data (online supplemental tables C and D). This was to assess whether meat imports decrease if fish becomes cheaper and is consumed more, or if rice acts as a potential staple substitute for potatoes, or if crackers, bongo (savoury snack) and sugar are potential substitutes or complementary foods for fruit and vegetables. Changes in retail prices of locally sourced foods were assessed when import tariffs on similar food groups were removed.

A change in tax revenue was calculated from combined excise, tariff and consumption tax revenue in the 12 months after each tax increase compared with the 12 months before.

For each indicator food, the waiver pass-through was estimated from the model derived decrease in retail price (T$/L) divided by the tax decrease (T$/L). The tax decrease was calculated by multiplying the size of the waiver (%) and the import value of the indicator food in the same year (T$/kg, Tonga Customs dataset). Tax waiver elasticities were estimated as the percentage change in import volumes for each percentage point tax reduction.

All analysis was done using R version 4.3.1 (R Statistical Foundation, Vienna, AT). GLS regression was fitted using the nlme package.²⁴

Author reflexivity statement

Tongan stakeholders met with the study team and supported the conceptualisation of the study research question, expressing their interest in the overall impact of excise tax policies on the healthiness of import foods, which is also reflected in Tonga’s non-communicable disease plan. Early career researchers who collated the data were funded for their time and included as co-authors, and collaborative meetings were used to discuss data quality improvements and offer further input on data cleaning. Preliminary analyses were shared in person at a key stakeholder meeting in Tonga in July 2024 and audience feedback at the meeting (and subsequently) informed improvements in the manuscript. Our team maintains relationships with key stakeholders and collaborators in Tonga which includes contributing to key Tongan non-communicable disease control meetings and health promotion training. The study team has trained Tongan nursing staff on research methods. We continue to use research funding to support Tongan colleagues, for example in drafting a policy brief. A New Zealand-based Tongan doctor was a key member of our small study team facilitating our relationships in Tonga.

Results

Seven food tax waivers were evaluated comprising three import tariff reductions, three import tariff waivers and a consumption tax waiver on fruit. The average percentage point waiver was 15%, with only one waiver that was larger (20%) and one that was smaller (10%).

Food prices after tax waivers

On average, there was a suggestion of only a small decline in retail price of waivered foods (−4%, 95% CI −10% to 2%) with a mean pass-through of 51% across foods and a median pass-through of 14%. Only two of nine food taxes demonstrated significant evidence of price declines: frozen vegetables (−20%, CI −25% to −14%) and cooking oil (−3%, CI −5% to 0%) (table 1, figure 1, online supplemental figure A). Price changes were small or negligible and not statistically significant for tinned mackerel, apples, oranges, potatoes and onions, except for a slightly larger price decrease for oranges after the 2016 consumption tax waiver (−17%, CI −34% to 5%). There was suggestive evidence of unexpected price increases for oranges in 2015 (10%, 95% CI −14% to 40%) and for frozen fish in 2013 (9%, CI −9% to 30%). If frozen fish was excluded from the meta-analysis (because of pre-existing tax concessions), the average decline in retail price of waivered foods was −5% (CI −11% to 1%). These price findings were robust to changes in model specifications (online supplemental table E).

Table 1. Impact of tax waivers on taxed food prices and import volumes, for the first-year post-waiver, Tonga.

Imported food product (indicator food in the price analysis)	Year	Tax waiver (% point)	Tax waiver9 (T$/kg)	Average price change (T$/kg, 95% CI)	% change in price (%, 95% CI)	Waiver pass-through (price change in T$/kg/ tax waiver in T$/kg) (%, 95% CI)	Import volume change (kg/person/year, 95% CI)	% change in imports (%, 95% CI)	Tax waiver elasticity (% change imports/tax waiver, 95% CI)
Fresh & frozen fish1 (frozen fish)	2013	15%	−0.36	0.91 (−1.01 to 2.83)	9 (−9 to 30)	−251% (−780% to 279%)	−0.1 (−4.8 to 1.8)	−4 (−79 to 338)	−0.3 (−64 to 23)
Tinned fish1 (tinned mackerel)	2013	15%	−0.76	0.03 (−0.22 to 0.29)	1 (−5 to 7)	−4% (−38% to 29%)	0.9 (−2.9 to 3.4)	35 (−48 to 248)	2.3 (−3 to 17)
Vegetable oils2 (cooking oil)	2013	10%	−0.22	−0.18 (−0.37 to 0.00)	−3 (−5 to 0)	82% (0% to 169%)	−3.1 (−10.9 to 0.6)	−50 (−79 to 18)	−5.0 (−8 to 2)
Fresh & frozen vegetables (frozen mixed vegetables)	2015	20%	−0.58	−1.67 (−2.26 to −1.09)	−20 (−25 to −14)	290% (189% to 392%)	0.4 (−1.0 to 1.2)	58 (−52 to 420)	2.9 (−3 to 21)
Fresh fruit3 (apples and oranges)	2015	15%	−0.30	−0.04 (−1.28 to 1.19)	−1 (−15 to 16)4	13% (−393% to 423%)	0.1 (−13.1 to 4.4)	1 (−73 to 276)	0.1 (−5 to 18)
			−0.64	0.94 (−1.27 to 3.14)	10 (−14 to 40)5	−146% (−489% to 198%)
	2016 (CT8)	15%	−0.31	0.07 (−1.42 to 1.55)	1 (−17 to 23)4	−23% (−505% to 462%)	−5.3 (−38.4 to 2.6)	−51 (−88 to 103)	−3.4 (−6 to 7)
			−0.43	−1.86 (−4.42 to 0.70)	−17 (−34 to 5)5	435% (−164% to 1034%)
Potatoes & onions (potatoes and onions)	2016	15%	−0.22	−0.23 (−0.66 to 0.20)	−7 (−20 to 7)6	104% (−91% to 299%)	2.0 (−7.1 to 1.0)	−32 (−63 to 26)	−2.1 (−4 to 2)
			−0.21	−0.03 (−0.52 to 0.46)	−1 (−15 to 15)7	14% (−216% to 245%)
Overall summary	2013–2016	15%	−0.40		−4 (−10 to 2)	Mean 51%, median 14%		−20 (−45 to 15)	−1.3 (−3 to 1)

The waiver comprised of existing import tariffs being replaced with zero tariff rates, except for: (1) Waivers on fish, which reduced import tariff to from 20% to 5% (note revenue collected from the 20% tariff was minimal), and then 0% subsequently in 2015. (2) Import tariffs on vegetable oils reduced from 20% to 10%. (3) On fruit, there was a 15% tariff waiver in 2015 and a 15% consumption tax waiver in 2016. (4) Apples. (5) Oranges. (6) Potatoes. (7) Onions. (8) Consumption tax waiver. (9) Tax waiver in T$/kg was calculated by applying the waiver (%) to the import value of the specific food (T$/kg) which was evaluated in the price model. This waiver (T$/kg) was used to calculate the waiver pass-through from the average price change for the indicator product.

Bold text indicates statistical significance, p<0.05.

Figure 1. Food retail price changes in the first year after each tax waiver, Tonga. Note: Interrupted time-series analysis findings were summarised by random effects (RE) meta-analysis. ‘Fruit 2016’ was a consumption tax waiver. All others are waivers of import tariffs. Data sourced from Tonga Department of Statistics.

Import volumes

There was no statistical evidence of an increase in import volumes of waivered foods (overall meta-analysis −20%, CI −45% to 15%), rather there was a suggested decline. Effects, however, varied by food. The largest observed increases in import volumes were for fresh and frozen vegetables (58%, CI −52% to 420%) and tinned fish (35%, CI −48% to 248%); and the largest observed declines in import volumes were for fruit after the consumption tax exemption in 2016 (−51%, CI −88% to 103%) and vegetable oils after the tariff reduction in 2013 (−50%, CI −79% to 18%) (table 1, figure 2, online supplemental figure B). In sensitivity analyses (online supplemental tables E and F), estimates for changes in fresh and frozen vegetable imports varied with different model specifications but remained within confidence limits of the main estimate. Placebo tests were generally supportive of the main results with findings in the opposite direction to observed waiver effects.

Figure 2. Import volume changes in the first year after each tax waiver, Tonga. Note: Interrupted time-series analysis findings were summarised by random effects (RE) meta-analysis. ‘Fruit 2016’ was a consumption tax waiver. All others are waivers of import tariffs. Data sourced from Tonga Ministry of Revenue & Customs.

If the fresh and frozen fish waiver (due to pre-existing tax concessions) was excluded from the meta-analysis, there was an almost equivalent average decrease in food imports (−21%, CI −46 to 15) as the overall result with all waivers included. If the consumption tax waiver was excluded from the meta-analysis (because it has a different mechanism), there was also a similar −17% (CI −44% to 21%) change in waivered food imports.

The average tax waiver elasticity indicated a (paradoxical) −1.3% (95% CI −3% to 1%) decline in import volumes for a 1% tax waiver. Elasticities ranged widely from 2.9 for fresh and frozen vegetables (where imports increased in response to the waiver) to −5.0 for vegetable oils (imports declined) and had wide uncertainty.

Potential substitution and complementarity

The average change in prices across potential substitute foods (to the tax waivered foods) was almost zero (−1%, CI −3% to 1%) (online supplemental figures C and D). The average change in import volumes across potential substitute foods was a suggested decline of −35% (CI −61% to 8%) (online supplemental figures E and F), that is, in the same direction as the import volume declines for tax waivered foods. There was a significant decline in import volumes of rice (−72%, CI −88% to −35%).

We hypothesised that the greater import increases after some tax waivers (tinned fish and fresh and frozen vegetables) might be affected by import prices being cheaper than locally sourced substitutes. Imported frozen vegetables (eg, mixed frozen vegetables) retailed at about half the average price of local vegetables, and imported tinned fish was also about half the price of locally sourced seafood. At the same time, imported potatoes and onions had a similar price to the average price for local tubers.

Revenue after tax waivers

Online supplemental table G shows the change in revenue (as a proportion of import values) for each tax waivered food. Forgone tax revenue to the Tongan Government totalled T$795 120 (US$330 200) in the first year after each tax waiver (approximately T$7.85 a person). Revenue changes were approximately the same size as the tax waiver, except for fresh and frozen fish where the tax intake was similar before and after the waiver (16% and 13%) due to the large proportion of pre-waiver tax concessions.²⁵

Discussion

Summary and interpretation

One of the criticisms of consumption tax waivers is that they will not be substantively passed on to consumers in terms of lower prices. The overall average price decrease of tax waivered foods in this study was 4% (but ranged from a 10% decrease up to a 2% price increase for specific foods). The average pass-through of tax waivers to reduced retail prices was 51% (median: 14%). The level of the import tariff waivers was further reduced for consumers, by the mark-up in value between imported food and its retail price. Our overall study findings are consistent with incomplete pass-through for the Tongan tax waivers (at 51% on average). This is consistent with low levels of pass-through (30%–60%) reported in other countries after reductions in consumption tax.^{13 26 27} For example, a 30% pass-through was observed after VAT decreases in European countries.¹³ It is likely that some of the savings from these reduced taxes have benefited importers, wholesalers and retailers, rather than consumers. Complete pass-through, however, has also been reported in some contexts.²⁸

The pass-through for the tax waivers did, however, vary markedly by food ranging from −251% on fresh and frozen fish to 435% for oranges. For frozen vegetables, the 20% import tariff waiver was followed by a 20% reduction in retail price (290% pass-through) and a 58% increase in import volumes of fresh and frozen vegetables. Pass-through larger than 100% may be a result of reduced imported food prices, increased competition and/or use of loss-leader pricing to increase market share. Also, in Tonga, the high level of pre-existing tariff concessions on fresh and frozen fish imports (for a variety of potential reasons)²⁵ meant reductions in tariff rates made little difference to price.

Perhaps reflecting the small price reductions at the retail level, study findings showed limited waiver effects on the levels of importing targeted food. There was an unexpected 20% average decrease in import volumes (ranging from a possible 45% decline to a 15% increase), and the extent of changes varied across targeted foods, for example with suggested increases in imports of fresh and frozen vegetables and tinned fish. Our findings contrasted with the evaluation of a tariff exemption on fruit and vegetables in Fiji which led to increased import volumes of these foods,⁴ but were consistent with findings from Tonga on the consumption tax exemption for fruit which had no flow-on effect to retail prices of apples and oranges.²⁰

Similarly, the waiver elasticities in this study were in an unexpected direction (averaging −1.3, CI −3 to 1). Waiver elasticities varied a lot by food, and all food specific confidence limits included plausible positive values. These findings contrast in direction with elasticities from meta-analyses of direct subsidies targeting fruit and vegetables, which indicate waiver elasticities of 0.59 for sales outcomes (an increase in sales) and 0.17 for consumption outcomes (increase in consumption, not statistically significant).⁹

The unexpected average decline in import volumes (and negative waiver elasticity) may have been affected by contextual factors. There appeared to be an overall shift away from imported foods to locally sourced foods during the study period¹⁹ (figure 2, online supplemental figure E). The broad range of excise taxes on unhealthy imported foods introduced in 2016 may have reduced the food budget available for other foods and thus encouraged the purchase of cheaper foods, which often may be those locally produced (and especially if home grown).

There were more favourable increases in import volumes for fresh and frozen vegetables (waiver elasticity 2.9) and for tinned fish (elasticity 2.3). Tinned fish and frozen vegetables (a proportion of the vegetable group) were cheaper than locally available substitutes like fresh fish and vegetables and are more convenient with a longer shelf life. Also, they come pre-packaged in fixed package sizes, which make it easier to ensure consistent pricing. Inconsistent pricing in foods and the low use of weighing scales in most shops has been cited as a reason for limited consumption tax exemption effects for apples and oranges in Tonga.²⁰

Strengths and limitations

This real-world natural experiment design used a robust interrupted time series analysis, adjusting for multiple potential time-varying confounders. Tonga’s tax waivers are a useful example to evaluate given (1) the middle-income country context, (2) the stated rationale to improve health, (3) waivers at levels averaging 15% which were expected to have a measurable effect, and (4) the inclusion of a range of healthy foods. The analysis was supported by obtaining perspectives from Tongan experts on local contextual factors that may have influenced study findings.

Some of the CIs presented for import changes are wide, with import volumes being small in magnitude and characterised by large monthly variations. The imprecision in estimates from individual food analyses was mitigated by inverse variance weighting in meta-analysis conducted to estimate the overall waiver effects. Price analyses were robust to variation in model specifications, but import estimates were more variable with adjustments to model specifications.

Customs import datasets required comprehensive cleaning. Any remaining randomly distributed measurement bias is expected to be non-differential. Other factors that may be affecting import trends (and possibly prices) over time are the Tongan food and beverage excises (broadly applied in 2016), other potential international food price trends, any Tongan promotion of locally sourced foods, and price controls (if they were followed which was not always the case).²⁰ Also, this analysis investigates 30+ outcomes at the 95% confidence level. Each association was analysed based on a priori specification and data collected for making specific comparisons, leading to fewer errors of interpretation.²⁹

Further research is needed to assess potential substitution or complementary changes in the supply of locally sourced foods, over and above the retail prices for these items reported here. Locally sourced fish, fruit and vegetables are widely available in Tongan markets and stores, but these were either not subjected to taxes or experienced no exemptions. Purchasing and consumption measures would be useful to investigate, including within subgroups such as age, income or rurality. Import supply is expected to be indicative of purchasing and consumption levels but does not correlate precisely due to potential variation in storage times and waste.

Implications

Pass-through is an important requirement to achieve intended health-promoting tax waiver effects. Overall, pass-through of tax waivers was moderate in Tonga, although the policy did appear to work well on some relatively cheaper and more convenient imported foods. The significant price drop for frozen vegetables was associated with a suggestive increase in import volumes of fresh and frozen vegetables and there was also a slight increase in tinned fish imports. Import volume findings should be interpreted cautiously considering the imprecision of model estimates and the contextual influences that may have reduced food imports. Possible influences include the broad range of excise taxes on unhealthy imported food largely since 2016, government actions to promote growing local staples and raising poultry, pigs and goats to substitute for imported foods, and other economic factors reducing food affordability. Supporting this, food inflation spiked in Tonga in 2017 and 2018,³⁰ and there was evidence of increasing domestic food production specifically for the staple crop of cassava (the main root crop).³¹

The 2016 tariff and consumption tax waivers did not appear to work any more favourably than earlier waivers, despite their timing coinciding with the introduction of a large number of excise taxes targeting turkey tails, mutton flaps, chicken leg quarters, instant noodles, mayonnaise and ice cream. However, this could not be definitively assessed in this analysis given the widely variable estimates of import changes after tax waivers across different years.

Evaluation of tax waiver effects in other settings, including with other implementation designs, is needed. Proposed policies should consider mechanisms to maximise pass-through to retail prices. It may be possible that other real-world tax waivers can achieve higher levels of pass-through than seen in Tonga.

Deleterious substitution to unhealthy imports was not observed in this study, but we could not rule this out given the general declining level of imports.

Conclusion

While there was no evidence of increased food supply overall for tax waivered foods, favourable price declines were observed for two of the nine foods (fresh and frozen vegetables and tinned fish). There was also evidence of moderate pass-through of tax waivers to lower retail prices. Nevertheless, interpretability was limited by highly variable estimates and a shift from imported to locally sourced foods during the study period. Further evaluation of tax waiver effects in other settings and with other implementation designs is needed.

Data availability statement

Data may be obtained from a third party and are not publicly available.