Site icon Converge

Should we use multi-year, moving-average of income to calculate tax & benefit payments?

Income tax rates are based on current/last year’s income. This makes them easy to calculate and implement.

This immediacy of taxes also makes them painful, and makes the tax slab thresholds as artificial barriers to income mobility. An example of this is when we get a raise which pushes us from near the top end of one tax rate bracket, to the bottom end of a higher tax rate bracket. This frequently means that even though the employer is paying us more after the raise, we are actually taking home less money due to a higher tax rate.

Government benefits work similarly. For example, the unemployment benefit / social support payments cut off (or reduce dramatically) when we start working. However, after accounting for taxes and loss of benefits, the take home income from pay is often lower than the unemployment benefits.

Some usual solutions to these issue from the right may be – single tax slab across income levels for the former, and reducing/removing unemployment benefits for the latter.

Some usual solutions from the left might be – more progressive tax slabs with higher min income thresholds and higher taxes on high income earners for the former, and a combination of higher minimum wages and slow tapering benefit programs for the latter.

I think there’s a third solution going untried. Both, income tax slabs and social benefit payments, should be based on a moving average income over 3-5 years, instead of the current/last year income.

This income smoothing (for purposes of tax rates) helps remove the sudden shocks to real take-home income from transitions between tax bands and benefit conditions. As a side effect, this smoothing would also help stabilise income tax receipts (and benefit payments) for the government – e.g. around steep bust boom years like 2005-20091.

For most of their history, tax rates and benefits had to be calculated manually (mentally?) by people. Complexity of income tax rates was loosely bound by computing power of pen, paper and an average human’s brain. Using moving-average income in this case would have made the job a fair bit harder. This was a sufficient deterrent from trying it. However, given current computing resources, this calculation is a much simpler job2. It may be time to remove the constraint of tools we no longer use, and design with the tools we use every day3.

There is already a version of this moving-average smoothing available to the rich and the corporations. They can claim tax credits through bad years, and use them to cut down on their taxes in good years. Using moving-average-income based taxation and benefits1 brings this stability to the more vulnerable in lower and middle-income classes as well.


  1. There seems to be one gap here. Benefits can be removed/reduced based on moving average income, they may need to be introduced based on immediate loss of income to prevent leaving receivers destitute if they suddenly lose their jobs.
    Similarly, while total loss of income wouldn’t impact for income tax rates (X% of 0 is still 0), a severe reduction in income could be damaging – a high tax rate, based on last few years’ income, on a now tiny income.
    Saving in good years, to pay taxes in case of bad years, is a way around this. But that’s in the domain of personal finance and behavioural economics, not tax policy. 
  2. It’s still not trivial. The complexity of tax codes and benefit programs ensure that nothing about them is trivial. Further, the constant political tinkering of them both means it’s not just not trivial, it might actually be a hard problem. But, I’m hypothesising: this is still a simpler problem than doing a single-year based calculation of today’s tax rates using just pen, paper, and an average person’s brain (the tools whose constraints they were designed around). 
  3. The smart amongst us may have another valid point here: Why just use tools of today, why not tools of tomorrow. Why use basic computing to calculate tax bracket based on a moving average, when we can use predictive modelling (“AI!!”) to fine tune and personalise the tax brackets even further?
    My concern here would be along explainability and predictability/verifiability. Unlike Uber fares, tax rates are a highly political topic. Politician would be wary of passing a bill where they have to sell the public to “trust” a black box algorithm to correctly predict their tax slabs (explainability).
    The second issue here would be predictability. The calculation method still has to be simple enough for someone to be able to calculate independently. This makes it verifiable, as well as makes it predictable – people can plan for it. 
Exit mobile version