I've decided to revisit this post from last year. The background context is the Conservative government's 'starve the beast' agenda - documented by me here and here, by Livio here and by Paul Wells in his excellent new book. The Conservatives have made two significant tax cuts during their time in office. The GST rate was reduced from 7% to 5%, and the corporate income tax (CIT) rate continued the downward trend begun during the Chrétien-Martin years: the statutory rate went from from 21% (22.1% with the surtax) in 2006 to 15% (no surtax) now. So far, the only tax cut anyone seems to want to rescind is the CIT. The NDP has already said as much, and the Liberals may or may not follow suit as they did in 2011.
So I was probably wrong to low-ball Canadian firms' propensity to reduce taxable income when tax rates increase. The unconditional correlation between federal CIT rates and federal CIT revenues (as a percentage of GDP) is actually negative.
The other thing is that had I skimmed through Bev Dahlby and Ergete Ferede's International Tax and Public Finance article too quickly: I didn't realise that they had broken out the effects of changes in provincial and federal CIT rates. (The reason for separating them is that since it's easier to shift income across provinces than it is to shift income in and out of Canada, firms are likely to react more to changes in provincial CIT rates than they do to federal rates). I'm going to use their model and estimates in what follows.
The basic model looks like
[log of tax base]t = η • [tax rate]t + λ • [log of tax base]t-1 + controls
This time, I'm also going to take account of the dynamics in that λ coefficient, especially since any sort of tax proposal in the next election campaign will include projections over a four-year horizon.
The Dahlby-Ferede estimates (standard errors) for η and λ are -1.657 (0.379) and 0.602 (0.084), respectively. I'll use 2013-14 corporate tax revenues as a base case (denoted by year 0 below): $34.986b from a tax rate of 15% implies a corporate income tax base of $233.24b.
If Δt is the change in the tax rate, then change in the log of the tax base over time relative to the base case looks like:
[log of tax base]1 = [log of tax base]0 + η•Δt
[log of tax base]2 = [log of tax base]0 + η•Δt(1 + λ)
[log of tax base]3 = [log of tax base]0 + η•Δt(1 + λ + λ2)
[log of tax base]4 = [log of tax base]0 + η•Δt(1 + λ + λ2 + λ3)
In the long run, the new tax base converges to
[log of tax base]∞ = [log of tax base]0 + η•Δt/(1-λ)
As a sort of robustness check, I'll look at parameter combinations that are the main estimates plus and minus one standard error:
- (η, λ) = (-1.657, 0.602) - base case
- (η, λ) = (-1.278, 0.518) - weak response case (effect reduced by one standard error)
- (η, λ) = (-2.036, 0.686) - strong response case (effect increased by one standard error)
Another thing to consider is the effect on provincial tax revenues. Since both levels of goverment apply taxes to the same base tax, changes in the tax base induced by a tax change at one level will affect the revenues of the other level. I'll use 11.1% as the provincial CIT rate, which is the weighted average estimated by the OECD.
And as a point of comparison, I'll also include the 'static analysis' results, in which it is assumed that firms do not react to higher rates by reducing the tax base.
Here are the effects of increase the federal CIT rate by 3 percentage points from 15% to 18%:
Revenue changes: Increase federal CIT rate from 15% to 18% Year Base case Weak response Strong response Static case Fed Prov Fed Prov Fed Prov Fed Prov 1 $5.0b -$1.3b $5.4b -$1.0b $4.5b -$1.5b $7.0b 0 2 $3.8b -$2.0b $4.6b -$1.5b $2.9b -$2.5b $7.0b 0 3 $3.1b -$2.4b $4.2b -$1.7b $1.8b -$3.2b $7.0b 0 4 $2.7b -$2.7b $4.0b -$1.8b $1.1b -$3.6b $7.0b 0 4-year sum $14.5b -$8.3b $18.3b -$6.0b $10.3b -$10.9b $28.0b 0 Long term $2.1b -$3.0b $3.8b -$2.0b -$0.4b -$4.6b $7.0b 0
In the base case, federal revenues start out as roughly 70% of the static analysis case, and converge to about 30% in the long run. By the fourth year, the increase in federal revenues is exactly offset by the reduction in provincial revenues.
What would happen if you increase the CIT rate by 6 ppts to bring it back where it was when the Conservatives elected?
Revenue changes: Increase federal CIT rate from 15% to 21% Year Base case Weak response Strong response Static case Fed Prov Fed Prov Fed Prov Fed Prov 1 $9.4b -$2.5b $10.4b -$1.9b $8.4b -$3.0b $14.0b 0 2 $6.8b -$3.8b $8.6b -$2.8b $4.9b -$4.8b $14.0b 0 3 $5.3b -$4.6b $7.7b -$3.3b $2.7b -$6.0b $14.0b 0 4 $4.4b -$5.1b $7.3b -$3.6b $1.2b -$6.8b $14.0b 0 4-year sum $25.9b -$15.9b $34.0b -$11.6b $17.1b -$20.6b $56.0b 0 Long term $3.2b -$5.7b $6.8b -$3.8b -$1.8b -$8.3b $14.0b 0
Revenues are higher with the larger increase, but they also decline faster. In the long run, federal revenues are less than 25% of the static case, and provincial losses are increased.
Finally, here's what happens when the federal rate is increased 9 ppts to 24%:
Revenue changes: Increase federal CIT rate from 15% to 24% Year Base case Weak response Strong response Static case Fed Prov Fed Prov Fed Prov Fed Prov 1 $13.2b -$3.6b $14.9b -$2.8b $11.6b -$4.3b $21.0b 0 2 $9.1b -$5.5b $12.0b -$4.1b $6.1b -$6.9b $21.0b 0 3 $6.8b -$6.6b $10.6b -$4.8b $2.7b -$8.5b $21.0b 0 4 $5.4b -$7.2b $9.9b -$5.1b $0.6b -$9.5b $21.0b 0 4-year sum $34.5b -$22.9b $47.4b -$16.9b $21.0b -$29.1b $84.0b 0 Long term $3.5b -$8.1b $9.1b -$5.5b -$3.8b -$11.4b $21.0b 0
Probably the most striking pattern is just how quickly the dynamic effects kick in: 80% of the adjustment to the long run is done by the fourth year. Another is how small the long-run effect on revenues is compared to what the static model predicts.
So the answer to the question in the title is "Not very much at all."