Income and Behavioural Responses at the Top of the Income Distribution

Elasticity of Taxable Income (ETI)

  • Measure that captures the full dimension of behavioural responses
  • Empirical quantity
  • Key for determining the efficiency cost of taxing top earnings + optimal top tax rates

Channels of Taxable Income Response:

  • Labour Income: Hours worked, participation, migration, effort on the job, training, education
  • Capital income: Savings, portfolio choice
  • Tax avoidance: Legal shifting of income into untaxed or lower-taxed form
  • Tax evasion: Illegal under-reporting of income

Top Income Share Analysis

Background in US:

  • At the end of WW2, top MTR was extremely high (94% in 1944-1955) and kicked it only at high incomes. Since then, MTR has fallen considerably to the 35% it is currently
  • Changes in the top MTR provides quasi-experimental variations that allows us to evaluate ETI among top-income earners
  • We evaluate responses by considering the change in top income share around the reform

Definitions:

  • Top income share = Total taxable income in the top percentiles of the distribution/share of total taxable income in the population
  • Taxable income = Reported before-tax income entering the calculation of tax liability

Insights

Observations

  • Top percentile share started to increase in 1981, when top MTR started to decline
  • Sharp jump in top percentile share from 1986-1988 corresponds to the sharp drop in top MTR enacted by TRA86
  • Top-percentile share continues to increase in the 1990s, despite increases in top MTR
  • Income share for the 90-99th percentile is very smooth, and displays no correlation with the MTR for this group

Conclusion: ETI may be strongly heterogenous across reforms/time periods. Alternatively, there could be many different confounders/factors muddying the results/conclusions here

Methodology

Top Income Share Analaysis

Full Variation over Time

Identifying assumption: Absent the tax change, top income share would have remained constant

DiD approach using whole population as a control group

Identifying assumption: Absent any tax changes, top income share would have remained constant or moved in a way that is uncorrelated with the top MTR.

Threat to identification: If the top income share is trending for non-tax reasons in a way that is correlated with the tax reform changes, this will bias the ETI estimate

ETI = change in ln(top income share)/change in ln(1-t)

Regression analysis: ln(top income share) = ETI*ln(1-t) + v

Trending Inequality

Regression analysis: ln(top income share) = ETI*ln(1-t) + at + bt^2 + ct^3 + v

Identifying assumption: Any non-tax changes in the top income share correlated with the tax reforms are fully captured by the 3rd order polynomial time trend controls

Threat to identification: We cannot be sure how to fully control for inequality trends, and sufficienly rich time trend controls will soak up all the identifying variation

Brewer-Saez-Shepard (2010), Kleven and Schultz (2014)

TRA86 Behavioural Response

Potential Bias:
If inequality increases for non-tax reasons, ETI estimate will be biased upwards. Other non-tax reasons include:

  • Non-tax policies favouring the rich
  • Skill-biased technical progress
  • Globalisation
  • Superstar markets
  • Changing social norms

We controlled for time trends and explored control groups very close to the top in order to rectify this, but we cannot be sure which specification adequately controls for a trending distribution.

Dependent on the nature of the tax system (e.g. broader tax base + strong enforcement can lead to a lower ETI)

Laffer Curve

A marginal change dt leads to

  • Mechanical revenue effect dM
  • Behavioural revenue effect dB
  • High-income Laffer rate, t*, is dR = dM + dB = 0

Laffer rate represents an upper bound on optimal top marginal tax rate.

  • It would not be wise to increase tax rates beyond the Laffer rate.

By formula,

  • t* = 1/(1+ETI x alpha)
  • Alpha = z/(z-z-bar) = average income/(average income above cut-off)
  • Alpha functions as a pareto parameter. Empirically, alpha appears to be rather fixed.

Implication for ETI

  • ETI is not a structural parameter. It depends on avoidance and evasion, which depends on the tax and enforcement system
  • If MTR > Laffer rate, you should i) reduce the MTR; b) reduce the ETI
  • Optimal policy depends on the social costs and benefits of either policy
  • Govt should have a favoured range of ETI estimates rather than just one estimate
  • Changes in personal taxable income could be driven by income shifting between corporate/personal tax bases
  • ETI estimate based on personal income data would then overstate the revenue and efficiency implications of behaioural responses

Main Finding

  • Pre-86, breakdown of income had very few S-corps, but S-corps are becoming increasingly more dominant.
  • This is because while regular corporations are subject to double taxation (i.e. pay taxes both at the corporate and then at the individual income level), S-corps are exempt from paying taxes at the corporate level :

Insights:

  • Top income share remains flat in the 1960 and well into the 1970s despite substantial tax cuts
  • About 1/3 of the surge in taxable income around TRA86 was driven by S-corporation income, which suggests shifting from C-corp to S-corp
  • Partnership income rises dramatically after TRA86 as partnership loss tax shelters are closed
  • Dramatic shit in the composition of top income from dividends to wage and S-corp (showing the shift from rentiers to working rich)
  • Dramatic secular growth in top wage income since the early 1970s, with short-term spikes around 1988 ad 1992 due to timing effects