Taxing the rich for redistribution – outcomes and implications
Agarwal, Qian, Yeung and Zheng (2022) investigate how a permanent income tax increase in Singapore impacted the affluent and elevated consumption for the low-income groups.
30 January 2023
According to latest findings by Oxfam International – an international confederation of NGOs dedicated to alleviating poverty – the world’s richest 1% has bagged “nearly twice as much wealth as the rest of the world put together over the past two years”.
While the confederation has called for more taxation on the ultra-wealthy, closer to home, the Singapore government had in February 2015 announced an increase in personal income tax on the country’s top earners, and subsequently implemented the policy in March 2017. The redistribution of wealth supported by this move was aimed at elevating the consumption levels of the lower-income population.
In their study “Taxing the rich to finance redistribution – evidence from a permanent tax increase in Singapore”, Agarwal, Qian, Yeung and Zheng (2022) investigate how such a fiscal policy had impacted both the affluent and low-wage earners, and its role in raising aggregate consumption throughout the economy.
Background
In his 2015 Budget Statement, then Deputy Prime Minister and Minister for Finance, Mr Tharman Shanmugaratnam announced an increase in income tax for the top 5% of the country’s earners, with marginal tax rates increasing between one and two percentage points for those whose annual incomes exceeded S$160K.
The revised tax regime would apply to income earned in 2016, and given that Singaporeans may start to pay income taxes in March of the following year after reporting their income, the additional tax levy took effect in March 2017.
Meanwhile, redistribution programmes – such as enhancements to the existing GST Voucher and Workfare Income Supplement schemes, and the introduction of the Silver Support Scheme – started ahead of the taxes, and were already in effect soon after the announcement.
Empirical design
As the new tax laws would only affect people earning S$160K or more per annum, this segment formed the treatment group. Wealth redistribution, on the other hand, was primarily targeted at those earning below S$100K per year. This meant that the segment earning between S$100K and S$160K neither suffered from the tax hikes nor gained from the redistribution programmes. Consumers in this bracket were thus singled out by the researchers to form the control group. The spending patterns of the control group proxied consumer behaviour in a counterfactual scenario where the tax laws had not been passed.
Using two data sets, each respectively from an international credit card company and DBS Bank, Singapore’s largest bank serving more than 80% of the local population, the researchers conducted propensity score matching (PSM) to pair treatment subjects with closely matched counterparts in the control group. Similarity in demographic characteristics – such as age, gender, ethnicity, marital status and housing type – minimises idiosyncratic differences between treatment and control subjects, and at the same time attenuates biases arising from such heterogeneity. The purpose of PSM was therefore to further isolate the effect of the independent variable (increased taxes) on the outcome variable (spending behaviour).
The matched sample derived from the credit card dataset had 2,165 individuals in the treatment and control group respectively, while the matched sample from the bank dataset comprised 10,431 distinct individuals in both treatment and control groups.
Methodology
The main mechanism driving analysis in Agarwal et al.’s study is the difference-in-differences (DID) method, which uses pre-treatment differences among observational groups as a benchmark, and thereafter compares these against post-treatment differences for further analysis.
The researchers first estimated the average monthly expenditures of the treatment and control groups, and thereafter estimated the “spending gap”, which is the difference of the averages. The spending gap was tracked across time, during the benchmark period and during the post-treatment periods, respectively following announcement and implementation.
A key tenet of the DID method is the parallel trends assumption, which posits that differences among groups under investigation would persist over time in the absence of intervention. Conversely, if parallel trends persist, then a reasonable conclusion could be that there was no intervention, or, in the context of the study, that the intervention had no effect.
In their analyses, Agarwal et al. found that pre-treatment parallel trends – namely, the spending gaps between the treatment and control groups – had failed to differ significantly following announcement and implementation of the new tax laws. The spending behaviour of the treatment group had closely tracked that of the control group, and given that the control group operated as a counterfactual where there were no new tax laws, the researchers therefore concluded that government intervention had no effect on the spending behaviour of the treatment group.
Findings
Baseline regression analysis on average monthly credit card expenditure showed that subjects in the treatment group did not change their spending habits significantly, both after the announcement in February 2015 and implementation in March 2017. Relative to the control group, results showed that treatment subjects increased their monthly credit card expenses by a mere 0.8 cents per additional dollar of tax after the announcement, and 0.5 cents per extra tax dollar after implementation. These figures were not significant enough to establish any correlation between the new tax laws and consumer spending, meaning that the researchers could not reject the null hypothesis, which posited that the increased taxes had no effect on expenditure.
Analyses on bank data likewise returned similar results. Relative to the control group, the treatment group decreased overall spending by a negligible average of 0.1 cents per additional tax dollar after implementation. Similarly, this result was insufficient to establish any correlation between fiscal intervention and spending, and the researchers again could not reject the null hypothesis of no spending response to extra income tax.
As the bank data contained more details, Agarwal et al. decomposed total spending into its various components, which include cash withdrawals, bill payments and card expenses. Across all the different types of expenditure, the changes in spending following implementation were statistically insignificant and economically trivial in the sense of comprising only a minuscule portion (<1%) of the subjects’ monthly expenses.
Accounting for heterogeneity in the treatment group
On the whole, the top income earners had appeared indifferent to the tax hikes. However, this belies the fact that the treatment group was internally heterogeneous. While some spend less, others spend more, and in effect net out each other.
To address the possibility that the overall group statistics had been too coarse, and that findings had glossed over differences within the treatment group, Agarwal et al. performed a serious of robustness tests on subsets of individuals in the treatment group who were more likely to cut their consumption in response to income tax hikes. After studying more sophisticated consumers who might be more rational, more indebted consumers and those whose gross income had remained stagnant or declined, the researchers found no evidence that these subgroups had curtailed their consumption in response to the negative income shock arising from higher taxes.
The effect of wealth redistribution
Though the tax hike came into effect from March 2017 onwards, wealth redistribution programmes were promptly rolled out after the government announcement in February 2015.
Statistically significant results from regression analyses showed that consumers earning below S$80K per annum increased their credit card expenditure by between S$15 and S$28 per month following the announcement, with the first half of 2014 as the benchmark period. After actual implementation in March 2017, those earning between S$20K and S$40K continued to increase credit card spending by between S$15 and S$30, relative to the same benchmark period.
These findings suggest that while the tax increase does not directly affect the rich population, it has a positive redistribution effect, directly raising the lower-income groups’ consumption.
Over 31 months after the announcement of the tax increase, and after discounting for the macroeconomic multiplier effect, the low-income group observed in the study increased their consumption by S$3.8 million or S$1.5 million per year, which accounted for 60% of the additional taxes raised from the high-income earners (=S$2.5 million).
Policy implications
Amid escalating income inequality and the socioeconomic fallout it may lead to, wealth redistribution ought to remain high on any government’s agenda. However, with rising inflation, slowing growth and increasing trade deficits, governments may be hesitant to tap on debt as a source of funding for social and economic development.
Singapore’s experience with raising money via tax hikes provides an alternative to debt funding, and findings revealed in Agarwal et al.’s study could be indicative of outcomes elsewhere when similar fiscal tools are deployed.
Authors:
Agarwal, Sumit is the Low Tuck Kwong Professor at the School of Business and Professor in the departments of Economics, Finance and Real Estate at the National University of Singapore.
Qian, Wenlan is the Ng Teng Fong Chair Professor in Real Estate and Professor of Finance and Real Estate at the NUS Business School.
Yeung, Bernard is the Stephen Riady Distinguished Professor in Finance and Strategic Management at the National University of Singapore Business School.
Zheng, Huanhuan is an assistant professor at the Lee Kuan Yew School of Public Policy, National University of Singapore.