A consumer company forecasting sales from a government cash payment should set one spending rate for lower-income customers and another for higher-income customers.
A spending rate here means the share of each dollar received that people spend on goods and services. The case for two rates comes from a working paper on financial windfalls, meaning cash that arrives without warning. The National Bureau of Economic Research and the Federal Reserve Bank of New York both issued it in October 2026. The New York Fed version says it presents "preliminary findings".
The authors studied lottery winners in one Canadian province, which they do not name. They linked the winners to credit bureau records, the files that lenders report to. Those records cover 11,703 winners from 2015 to 2020. Each winner in the study won a single lump-sum prize of $1,000 to $250,000.
The paper measures spending from credit card records, scaled up to total spending with surveys from the Bank of Canada and Statistics Canada. It then sorts winners into a lower-income half and a higher-income half, by the average household income of each winner's neighborhood.
Winners from lower-income neighborhoods spent about 25 cents of each dollar won over two years. Winners from higher-income neighborhoods spent about 15 cents. The paper says high-income winners put a larger share into "savings and leisure".
The authors then apply the rates to a government payment. In their exercise, modeled on Canada's 2025 middle-class tax cut, each adult below median income gets one payment of a thousand Canadian dollars. A single average rate blends both groups equally, at 20 cents per dollar. That average predicts $3.42 billion of extra spending over two years. The lower-income rate predicts $4.32 billion, 26 percent more.
In a second version, only the lowest-income quarter of adults gets the payment. There, a rate specific to that group's income predicts 76 percent more spending than the single average. The authors say the single average "would underestimate the true consumption impact".
The evidence has limits for any company. A lottery prize is a surprise. The paper says comparisons across studies "can be complicated" by whether a cash gain is "anticipated or unanticipated". The authors also warn that responses measured in one income group "need not extrapolate to another".
The two-rate forecast runs on the company's own sales data, and its own finance and analytics teams do the work.
What to do first:
- Chief financial officer: before the next forecast that counts on a government payment, split customers into two groups by neighborhood income.
- Head of sales planning: give each group its own spending rate, starting from the paper's 25 and 15 cents per dollar over two years.
- Head of analytics: each quarter for two years after the payment, compare each group's actual sales with its forecast and reset the rates.
The Social Sciences and Humanities Research Council of Canada funded the research. The authors work at the Federal Reserve Banks of New York and Philadelphia and at universities in Canada and the United States. Each of the eight authors declares no relevant or material financial interest in the research.
