Taxing Top Wealth: Migration Responses and their Aggregate Economic Implications
What this paper finds — and why it matters
Layer 1: Overview
Research question and motivation: Proposals to tax top wealth (e.g., Saez and Zucman, 2019) face a recurring objection in public debate: that the wealthy will emigrate en masse and, because many are entrepreneurs, their departure will inflict large negative spillovers (“trickle-down”) on the broader economy, making wealth taxes self-defeating. Credible evidence on international migration responses to wealth taxes has been scarce due to data limitations and a lack of clean identifying variation. This paper provides such evidence and quantifies the aggregate economic implications.
Data and setting: The authors use exhaustive administrative data from Sweden (wealth tax register Förmögenhetsregistret 1993-2007, LISA, matched employer-employee RAMS, K10 closely-held-business filings, and the Serrano ownership-network data that maps indirect ownership) and Denmark (used for out-of-sample validation). A key strength is observing all wealth components without top-coding and linking individuals to firms they control directly and indirectly. They exploit three large reforms: the unexpected 2007 repeal of the Swedish wealth tax (statutory top marginal rate fell from 1.5% to 0%; effective average rate on the top 2% was ~0.5%), and Danish reforms of 1989 (rate cut from 2.2% to 1%) and 1996/1997 (abolition). Business assets were exempt in Sweden but fully taxed in Denmark.
Empirical strategy: A two-step procedure. Step 1 estimates migration elasticities using difference-in-differences around the reforms (treated = top 2% of net wealth; baseline control = top 20% to top 10%), with treatment assigned on predicted wealth to avoid endogeneity post-2007. Step 2 estimates the effect of migration on individual-, firm-, and market-level outcomes via event studies (never-movers with placebo dates as controls), independent of the tax reforms. The two are combined, weighted by the wealthy’s share of aggregate activity (decomposition in equation 1).
Main quantitative findings: A 1pp increase in the top wealth tax rate raises the out-migration rate by 0.17pp and reduces in-migration by 0.05pp; the 2007 repeal cut wealthy out-migration propensity by ~30% (about one-third of top-2% expatriations were tax-induced). Danish elasticities are statistically indistinguishable. Net flow semi-elasticity is -0.22pp per 1pp. Flow effects cumulate to a modest stock elasticity: the elasticity of the wealthy population w.r.t. the net-of-tax rate is 1.77 (s.e. 0.47) — a 1% rise in the net-of-tax rate raises the stock by under 2%. The implied income-net-of-tax migration elasticity is ~0.05, comparable to top-income cross-border elasticities. Firms controlled by the top 2% account for ~9% of Swedish employment, 15% of value added, 12% of investment, 19% of tax payments (and ~10% employment / 15% value added per the intro). When a top-2% owner out-migrates, directly-controlled firms see employment fall ~33%, gross investment ~22%, value added ~34%, and tax payments ~51%, driven almost entirely by the extensive margin of firm disappearance (effects near zero conditional on survival). But 45% of “closed” firms are absorbed via mergers/acquisitions; displaced workers lose only 4.3% in earnings and face a 0.6pp higher unemployment probability; market-level spillovers are small and insignificant even for granular firms.
Aggregate and policy implications: Combining steps, a 1pp rise in the top wealth tax rate reduces aggregate employment by 0.022%, investment by 0.065%, and value added by 0.103% in the long run — modest despite the wealthy’s large economic footprint, because migration flows are small. Fiscally, each $1 raised loses only $0.22 to migration responses vs. $0.54 to intensive-margin responses (savings/avoidance/evasion, using Jakobsen et al. 2020), so $0.76 total. Migration responses are far from the Laffer bound but, because the MCPF is highly nonlinear, they nearly double it from ~2.2 to ~4.2. Migration threats, while salient in debate, matter less for welfare and policy than intensive-margin responses.
Layer 2: Deep Dive
What is the identification strategy for the migration elasticity and what are the main threats?
A difference-in-differences design around the 2007 Swedish wealth tax repeal, comparing out-migration of the treated top-2% group to a control group in the top 20% to top 10%. The non-contiguous control avoids contamination bias (households near the threshold anticipating future liability; less than 1% of controls reach the top 2% by 2006). The main threat is the parallel-trends assumption given a control group lower in the distribution; the authors show no differential pre-trends in out-migration and that effective capital-income and labor-income tax rates evolved similarly across groups (only wealth-inclusive tax rates diverged). The 2007 inheritance tax abolition is ruled out as a confounder because inheritance tax had little bite and strict residency rules made it hard to avoid by migrating (10-year non-residence required at death). Treatment is assigned on predicted wealth (from pre-reform variables) to avoid endogenous post-2007 wealth measurement. 2SLS specification (4) instruments the log net-of-tax rate with the treatment-by-post interaction.
How is the aggregate effect identified separately from the migration channel, and why not use the reform directly?
National wealth tax reforms cannot identify general-equilibrium/aggregate effects because treatment and control groups share the same aggregate economy, the exclusion restriction fails (wealth taxes also affect savings, capital accumulation, avoidance/evasion), and they are underpowered (small stock changes are hard to detect). The two-step procedure circumvents this: event studies of migration events (specification 7, with randomly-assigned placebo dates for never-movers, no matching) give the effect of migration on outcomes independent of the tax reform, and these are combined with the reform-based migration elasticity, weighted by the wealthy’s share of each aggregate outcome (equation 1).
What is the role of the LATE / marginal-mover correction?
The two-step procedure requires the population whose migration impact is measured (event studies) to match the population whose migration responds to the tax (compliers). Using methods from the insurance-selection literature (Hendren et al., 2021) and the fact that 30% of pre-reform wealthy migrants were tax compliers, they recover the characteristics and treatment effects of marginal movers. Tax-induced movers (compliers) are slightly younger, slightly more likely entrepreneurs, slightly wealthier, around the 65th-70th skill percentile, but their firms are not selected. Event-study estimates pre vs post reform are similar (not statistically different), so treatment-effect heterogeneity is limited; column (5) double-difference LATE estimates for compliers are the preferred inputs.
What is the firm-level evidence and how is reallocation distinguished from genuine destruction?
Owner out-migration causes a ~30pp drop in firm survival (firm-identifier disappearance) and large declines in employment (~33%), value added (~34%), investment (~22%), turnover, and tax payments (~51%), almost entirely extensive-margin. The authors distinguish destruction from reallocation using Bolagsverket merger/closure-reason data: 45% of closures are linked to mergers (the firm is absorbed), 55% are liquidations/bankruptcies. Accounting for buy-outs cuts the firm-existence and employment effects by ~40%. Worker-level event studies show displaced employees lose only 4.3% in earnings and 0.6pp higher unemployment, indicating workers reallocate. Including indirectly-held firms, five-year effects are employment -19%, value added -33%, turnover -28%, investment -19%, tax payments -45%.
What heterogeneity is documented?
Migration semi-elasticities do not vary much by age or education; entrepreneurs’ out-migration semi-elasticity is larger but less precisely estimated (their effective tax rate dropped less because business assets were exempt; their out-migration fell ~0.14pp, roughly 50%, within a year). Firm-level migration effects show limited heterogeneity by owner age or children; effects are smaller for larger firms and especially for the top-10 largest moves (multi-billion-SEK businesses), where effects are considerably below average. In-migration effects mirror out-migration with opposite sign but are smaller for value added, turnover, investment, and tax payments.
What robustness checks are run?
Estimates are robust to alternative control groups closer to the treatment group; to assumptions on the regeneration/replacement rate of the wealthy population and to dynastic effects (detectable but small); and to tax evasion — using Alstadsæter et al. (2019) and Boas et al. (2024) bounds, the stock elasticity ranges 1.85 (lower) to 1.92 (upper) vs. 1.77 baseline. Firm outcomes are robust to winsorization choices (Appendix Table IV.3); with no winsorization, value added/investment/tax effects turn positive-insignificant due to one outlier firm. Market-level spillovers are insignificant across alternative market definitions. Alternative aggregate calibrations (including accounting for buy-outs) imply smaller effects, so the baseline is a conservative upper bound.
How does the paper relate to and differ from prior work?
It builds on the wealth-tax behavioral-response literature (Seim 2017; Jakobsen et al. 2020; Brülhart et al. 2022) which is largely silent on international migration, and on the tax-migration literature (Kleven et al. 2013/2014/2020; Akcigit et al. 2016) which focuses on income taxes and within-country mobility. It is the first systematic evidence on international migration responses to wealth taxes and their trickle-down. Versus the CEO/owner death-and-retirement literature (Smith et al. 2019: -26pp firm survival, -82% profits per worker, -45% even conditional on survival; Jäger and Heining 2022), migration effects are much smaller and nearly zero conditional on survival, because owners often retain control or restructure rather than shut down. Findings echo Bach et al. (2023) for France.
What are the policy implications and their scope conditions?
Migration-driven fiscal externality is $0.22 per $1 raised, vs. $0.54 for intensive-margin responses, $0.76 combined — below the Laffer bound. Because the MCPF is nonlinear, migration roughly doubles it from ~2.2 to ~4.2; wealth taxation would be welfare-improving if revenue funds projects with MVPF above 4.2 (e.g., programs for low-income children, often above 5 per Hendren and Sprung-Keyser 2020). Scope conditions: estimates come from reforms that only cut rates, so asymmetric responses to increases cannot be ruled out; the elasticity depends on destination-country taxes (Swedish movers went to low-tax UK non-dom, Switzerland, Austria), so responses could be more muted if all neighbors taxed wealth heavily; results are for small open economies with low wealth inequality and weaker agglomeration than the US, suggesting the estimates are upper bounds; computations reflect 1990s-2000s Scandinavia where offshoring/evasion mattered, and depend on tax base, enforcement, and exit-tax design.
How is the stock elasticity derived from flow elasticities?
Using a simple OLG framework, the population stock elasticity ≈ net-flow semi-elasticity times (T+1)/2, where T is the average ’lifespan’ of wealthy individuals (the inverse of the regeneration/birth rate into the wealthy population). Longer lifespan means slower regeneration, so lost migrants are harder to replace and the stock effect is larger. This yields a stock elasticity of 1.77 (s.e. 0.47); the effect stays modest because top-of-distribution migration flow rates are very small.
What are the magnitudes of migration flows and tax-payment effects, and any caveats on persistence?
Top-decile out-migration is ~0.2% per year in Sweden (vs. ~0.65% in the bottom half) and ~0.1% in Denmark, rising in the extreme tail; taxable wealth of wealth-tax-liable out-migrants is only 0.09% of total taxable wealth; net migration is small and slightly positive. One year after out-migration, total tax payments fall ~66% (wealth tax -59%, income tax -68%; income taxes are ~90% of the wealthy’s payments, implying large fiscal externalities on income tax). Effects attenuate over time: ~40% reduction at five years because ~40% of out-migrants return within five years (migration is persistent but return migration is common). Taxable wealth in Sweden falls 94% one year out; real estate is typically sold, and financial wealth falls at extensive (-21%) and intensive (-15%) margins, confirming real rather than purely fiscal-residence responses.