@article{LI2026, 
author = {Bing LI and David Daokui Li and Ke'aobo LI},
title = {How Many Hours Do We Work for the Government Each Year—A Theory of Time Tax Rate Stability and Its Implications for Government Size in the Age of Artificial Intelligence},
year = {2026},
journal = {Research in Government and Economics},
volume = {2},
number = {2},
pages = {116-159},
keywords = {time tax rate, tax consent, government size, labor productivity, artificial intelligence},
url = {https://www.sciopen.com/article/10.26599/RGE.2026.9720207},
doi = {10.26599/RGE.2026.9720207},
abstract = {Conventional theories often suggest that economic development is accompanied by a sustained increase in the tax-to-GDP ratio and a heavier tax burden on households. This paper reconsiders that link from the perspective of household time. To capture the labor tax burden as experienced by households, we introduce the concept of the time tax rate, which combines the effective labor tax rate with the share of annual waking hours devoted to work. This measure indicates the share of a worker's annual waking hours that can be interpreted as “working for the government.” Using panel data for 15 OECD countries from 1950 to 2019, we show that Tax/GDP rose from 24.8% to 36.5% and the effective labor tax rate from 12.7% to 30.7%. By contrast, the time tax rate remained substantially below both conventional measures, rising only from 4.6% to 8.2%, and entered a relatively stable plateau after the mid-1970s. In terms of eight-hour workdays, the time spent “working for the government” increased from about 34 days per year to about 60 days. This increase is not negligible, but it is considerably smaller than what the rise in the effective labor tax rate would suggest. We explain this pattern by arguing that higher labor productivity raises output per unit of time while also allowing workers to reduce working time, thereby cushioning the effect of higher tax rates on the household time burden. Theoretical analysis characterizes this mechanism, and panel regressions show that the long-run relationships among the relevant variables are consistent with it. Building on this framework, we further suggest that, in the age of artificial intelligence, Tax/GDP may continue to rise; yet if productivity gains from AI translate into higher labor income and shorter working hours, the time tax rate need not increase substantially and may even decline. The paper thus provides a household-time perspective for understanding both the sustainability of high-tax equilibria and the fiscal space created by artificial intelligence.}
}