Editors’ note: This column is based on CEPR Discussion Paper 21672 “Capital taxation, income shifting, and retained earnings: Evidence from dividend tax reforms”.

Governments around the world increasingly rely on capital taxation to raise revenue and reduce inequality. Temporary tax holidays on dividends, repatriated profits, or accumulated earnings are especially attractive because they promise large short-run revenue gains without permanently changing tax rates. Similar policies have been used in the United States, Israel, and several European countries, and proposals for one-off capital tax measures continue to appear whenever governments face fiscal pressures.

But do temporary tax incentives merely change when taxpayers realise income, or do they permanently change how they behave?

Evidence from our recent study of two dividend tax reforms in Israel suggests that the answer depends critically on the design of the reform (Berman and Klor 2026). Temporary tax relief may generate impressive short-run revenue gains, but it can also change taxpayers’ expectations in ways that encourage them to retain earnings and postpone future distributions, reducing tax revenues long after the tax holiday has ended.

Why dividend taxation is different

The taxation of dividend income has become increasingly important as capital income has grown relative to labour income. The debate over how best to tax wealthy individuals increasingly centres on whether governments should tax wealth directly or instead tax the income generated by wealth (Bastani and Waldenström 2024).

Dividend income occupies a particularly important place in this discussion. It is highly concentrated at the top of the income distribution – in the UK, dividends account for more than one quarter of total income for the top 1% and more than one third for the top 0.1% (Joyce et al. 2019). At the same time, dividend income is unusually flexible because owner-managers of closely held firms often decide when to distribute profits and when to retain them inside the company.

Previous research has consistently shown that dividend distributions respond strongly to tax changes while real investment changes comparatively little (Chetty and Saez 2005, Yagan 2015, Bach et al. 2024, Miller et al. 2024). Much less attention, however, has been paid to whether the design of a reform – temporary versus permanent – changes taxpayers’ behaviour in the years that follow.

Two dividend tax reforms

Israel provides an interesting setting to study this question because it implemented two very different dividend tax reforms within five years.

The first reform permanently increased dividend tax rates beginning in 2012. Anticipating higher future taxes, shareholders accelerated dividend distributions into 2011. The second reform moved in the opposite direction. During most of 2017, the government temporarily reduced dividend tax rates to encourage shareholders to distribute retained earnings from profits accrued by the end of 2016. Once the temporary window closed, tax rates immediately returned to their previous level.

These two reforms produced remarkably similar short-run responses but dramatically different longer-run outcomes.

Large short-run responses

Both reforms generated enormous increases in reported dividend income. Before the permanent tax increase took effect, dividend income more than doubled as taxpayers accelerated distributions to avoid higher future tax rates. Five years later, the temporary tax holiday generated an even larger response, with reported dividend income more than tripling during 2017.

The combined effect of the two reforms amounted to roughly 3–4 percentage points of national income in additional dividend distributions (Figure 1). Dividend tax revenue also increased sharply, reaching almost 4% of total tax revenues during the 2017 reform.

These responses were concentrated almost entirely among the very top of the income distribution. At the same time, wage income fell among top earners during the reform years, indicating that taxpayers temporarily relabelled labour income as dividend income to take advantage of more favourable tax treatment.

These short-run responses are broadly consistent with previous studies of dividend taxation.

Figure 1 Dividend income and dividend tax revenue in Israel, as a share of national income, 2005–2020

Notes: Each reported series is compared with a counterfactual that interpolates over the reform years (2011 for dividend income, 2012 for tax revenue, and 2017). Red arrows show the gap, in percentage points of national income. 
Source: Berman and Klor (2026).

The surprise came afterwards

The real difference between the two reforms emerged after the reforms ended. Following the permanent 2012 tax increase, dividend payments quickly returned to their pre-reform level. The reform largely brought forward distributions that would have occurred anyway.

Following the temporary 2017 tax cut, however, dividend payments remained roughly 55% below their pre-reform level for several years. Why?

One possibility is mechanical stock depletion. Firms may simply have exhausted accumulated retained earnings during the temporary tax holiday and therefore had less available to distribute afterwards. Our evidence suggests that this explanation cannot fully account for the observed pattern.

The permanent 2012 reform actually triggered a larger withdrawal of accumulated profits than the temporary 2017 reform. If depletion were the main mechanism, dividend payments should have recovered more slowly after 2012 than after 2017. Instead, the opposite occurred.

The evidence is more consistent with a change in expectations. Once taxpayers observed a temporary tax holiday, they may have begun expecting similar relief in the future. Rather than paying dividends at the normal tax rate, they retained earnings inside their firms while waiting for another opportunity to distribute them at a preferential rate.

Although stock depletion probably contributed to the initial decline, our evidence suggests that changes in expectations also played an important role in explaining the persistence of lower dividend payments after 2017.

Tax policy also changes measured inequality

The behavioural response has an implication that is rarely discussed. Retained earnings are generally not recorded as personal income until they are distributed. Consequently, tax-based measures of top incomes already understate inequality whenever firms retain profits rather than distribute them (Alvaredo et al. 2020, Smith et al. 2022).

The temporary tax relief made this problem worse. Because top-income taxpayers retained more earnings after 2017, their reported personal income fell even though their underlying economic resources did not fall by nearly as much.

Measured top income shares therefore declined substantially following the reform. Using a counterfactual based on pre-reform behaviour, Figure 2 shows that the 2017 reform itself increased the downward bias in the measured top 1% income share by roughly 1–2 percentage points. In other words, the design of a tax reform changed not only taxpayer behaviour but also the inequality statistics policymakers rely on to evaluate tax policy.

Figure 2 The top 1% income share in Israel: Reported income versus a no-reform counterfactual

Notes: The top 1% share is total gross income of the top 1% divided by total reported income. The counterfactual adds back the dividend income estimated to have been retained because of the 2017 reform. The reform years 2011 and 2017 are excluded. 
Source: Berman and Klor (2026).

Lessons for tax policy

Two broader lessons emerge. First, the design of capital tax reforms matters as much as the tax rates themselves. Temporary tax holidays may generate substantial short-run revenues but also encourage taxpayers to postpone future distributions if they begin expecting similar relief again. Policymakers evaluating one-off capital tax measures should therefore consider not only immediate revenues but also how today’s policy shapes tomorrow’s expectations.

Second, tax policy influences the measurement of inequality as well as inequality itself. When capital income increasingly accumulates inside firms rather than being distributed, tax records become a less reliable guide to the true distribution of income. Temporary reforms can widen that gap by changing reporting behaviour without producing equivalent changes in underlying economic inequality.

The broader lesson extends beyond dividend taxation. Whether governments are considering temporary capital tax relief, one-off wealth levies, or broader reforms to capital income taxation, policy design affects not only how much revenue is collected today, but also how taxpayers behave tomorrow – and even how we measure the success of the policy itself.

References

Alvaredo, F, A Atkinson, L Chancel, T Piketty, E Saez and G Zucman (2020), “Distributional National Accounts (DINA) guidelines: Methods and concepts used in the World Inequality Database”, World Inequality Lab.

Bach, L, A Bozio, A Guillouzouic, C Leroy and C Malgouyres (2024), “Follow the money! Why dividends overreact to flat-tax reforms”, mimeo.

Bastani, S, and D Waldenström (2024), “Taxing the wealthy: The choice between wealth and capital income taxation”, VoxEU.org, 5 March.

Berman, Y, and E F Klor (2026), “Capital taxation, income shifting, and retained earnings: Evidence from dividend tax reforms”, CEPR Discussion Paper 21672. 

Chetty, R, and E Saez (2005), “Dividend taxes and corporate behavior: Evidence from the 2003 dividend tax cut”, Quarterly Journal of Economics 120(3): 791–833.

Joyce, R, T Pope and B Roantree (2019), “The characteristics and incomes of the top 1%”, Institute for Fiscal Studies.

Miller, H, T Pope and K Smith (2024), “Intertemporal Income Shifting and the taxation of business owner-managers”, Review of Economics and Statistics 106(1): 184–201.

Smith, M, D Yagan, O Zidar and E Zwick (2022), “The rise of pass-throughs and the decline of the labor share”, American Economic Review: Insights 4(3): 323–340.

Yagan, D (2015), “Capital tax reform and the real economy: The effects of the 2003 dividend tax cut”, American Economic Review 105(12): 3531–3563.