Aggregate Studies

The Problem with Empirical Macro 1/3

The Problem with Empirical Macro 2/3

The Problem with Empirical Macro 3/3

Messerli (2012): Nobel Laureates and Chocolate Consumption.

The Problem with Empirical Macro

Why don’t we just look at what happens to GDP following a tax cut?

  1. Many things happen in any one year:
    – September 15, 2013: Britney Spears happens to release her new song, which promotes work ethic.
    – However, other things have happened between 2013 and 2018 (including a massive tax cut plan)

  2. Policies are changed for a reason:
    – Years where taxes are changed are different from taxes where taxes are not changed.
    – This is not like a Randomized Control Trial (RCT) in medicine: macroeconomic policies are not changed randomly.
    – For example: \(\Delta G>0\) often happens during recessions. Low subsequent GDP growth: low multipliers or because GDP growth was low to start with.

Answers

What are the potential answers?

  • Add up many tax changes:

    • Some tax changes are accompanied by a new release of Britney Spears, but on average they are not.

    • Allows to control for other types of more serious events, too. (wars, etc.)

  • State their motivations:

    • Taxes raised to reduce the deficit, or increase long-run incentives are “exogenous.”

    • We do not want to look at tax changes which are made for managing the business cycle, in particular.

List of Tax Changes 1/2

List of Tax Changes 2/2

Tax Increase, 1% of GDP (Romer and Romer (2010))

Romer and Romer (2010)

Advantages and Disadvantages

Issues with these studies:

– Noisy results: multiplier is between 2 and 4.

– One cannot further decompose: e.g. Top 10% VS Bottom 90%. We would get something even noisier.

– Always worry that tax changes are endogenous. (at the aggregate level, taxes are changed for a reason)

Advantages:

– It is exactly the object of interest (national level multiplier).

– Allows to tell apart different models.

Individual-Level Studies

Advantages and Disadvantages

Using individual-level data like survey, fiscal, administrative, or account-level data to measure \(\epsilon\), or \(c_{1}\). Advantages:

– Many more individuals: less noisy results (more observations).

– More credible “identification”: comparing two people at the same time period.

Disadvantages:

– Keynesian, aggregate demand effects cannot be estimated.

– e.g. if I decrease someone’s tax rate, then it might lead someone else to work more, not just the person who benefited from the fall in tax rates. (though the aggregate demand effect).

– Thus, there is no clean “control” group if there are aggregate demand effects.

MPC (Jappelli and Pistaferri (2014))

MPC (Jappelli and Pistaferri (2014))

Cross-sectional Studies

Advantages and Disadvantages

Identification across zipcodes, counties, or states. Advantages:

– More observations.

– Measure Keynesian, aggregate demand, general equilibrium effects.

– Less endogenous changes than at the national level: aggregate taxes are not changed in the U.S. to target California’s GDP specifically.

Disadvantages:

– Openness \(m_{1}\) of a state is larger, so multiplier is lower.

– But we are interested in national level multipliers, not state level multipliers.

– We thus need economic theory in order to infer national multipliers from state multipliers.

Zidar (2019): “Tax Cuts for Whom?”

Using state-level variation in income distributions, Zidar (2019) in a Journal of Political Economy paper named “Tax Cuts For Whom? Heterogeneous Effects of Income Tax Changes on Growth and Employment” estimates the following effects on GDP:

– Multiplier effect of a tax cut to the bottom 90% is roughly 7.

– Multiplier effect of a tax cut to the top 10% is roughly 0.

– A tax cut going half to both groups has a multiplier of about 3.5 (Romer, Romer (2010) result).

Zidar (2019): “Tax Cuts for Whom?”

Results seem to confirm our results from Lecture 9: tax cuts on bottom 90% work better than on top 10%.

Effects on employment are similar:

– 1% of state GDP tax cut for the bottom 90% results in 3.4% employment growth over a 2-year period.

– 1% of state GDP tax cut for the top 10% is 0.2% and statistically insignificant.

Zidar (2019)

Zidar (2019)

Zidar (2019)

Zidar (2019)

Zidar (2019)

Conclusion

Additional Watching (Not Required)

Taking Stock

  • Value of the Keynesian multiplier is still subject to very intense debates.

  • Tax-based multipliers > government spending multipliers (because tax changes are more persistent?)

  • Tax-based multipliers could be as high as 3.

  • Evidence that tax changes have long-term effect. There is a paradox of thrift.

  • My view: the evidence is more supportive of the Keynesian model, than of the neoclassical model.

  • Disclaimer: not everyone agrees with that view, and you are perfectly free to disagree too!

Bibliography

Jappelli, Tullio, and Luigi Pistaferri. 2014. “Fiscal Policy and MPC Heterogeneity.” American Economic Journal: Macroeconomics 6 (4): 107–36. https://doi.org/10.1257/mac.6.4.107.

Romer, Christina D., and David H. Romer. 2010. “The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks.” American Economic Review 100 (3): 763–801. https://doi.org/10.1257/aer.100.3.763.

Zidar, Owen. 2019. “Tax Cuts for Whom? Heterogeneous Effects of Income Tax Changes on Growth and Employment.” Journal of Political Economy 127 (3): 1437–72. https://doi.org/10.1086/701424.