RDP 2018-07: The GFC Investment Tax Break 2. Related Theory and Literature

This section reviews the economic theory regarding the determinants of business investment and outlines how the tax break we study relates to this theory.

The modern study of the relationship between business taxation and investment began with Hall and Jorgenson (1967), who derived an expression for the cost to a company of employing fixed assets:

This user cost of capital is a key determinant of investment choices in a standard investment model. P is the real price of investment goods and M is the weighted average cost of funds (debt and equity) of the company. τ is the company tax rate. The user cost for an unincorporated business replaces τ with the average marginal tax rate of its owners. Z is the ratio of the present value of future depreciation allowances to the initial purchase price of the asset. These allowances permit businesses to deduct the cost of the capital investment from their taxable income over a multi-year period that approximates the economic life of the asset.

Given discounting of future cash flows, in normal times Z is always less than 1. For equipment in Australia, a reasonable baseline value is 0.85. The investment tax break we study involves an additive increase to Z:

The early literature (including Hall and Jorgenson) used aggregate time-series data to show that the user cost was a significant determinant of investment. This early literature is well surveyed by Hassett and Hubbard (2002). While there was some apparent success, later contributions showed that most of the apparent relationship was due to the relationship between recent economic growth and current investment: this is the standard ‘accelerator effect’ that has been known about since Clark (1917). Cockerell and Pennings (2007) provide an Australian version of this result. In their models, the cost of capital generally has a statistically significant negative coefficient, but most of the explanatory power is provided by other cyclical variables such as business confidence and the real exchange rate.

The more recent literature has focused on using changes in tax parameters to estimate the relationship between the cost of capital and investment. Such changes are more plausibly exogenous than variation in financing costs (Cummins, Hassett and Hubbard 1994), and are less prone to measurement error. This is the approach taken in the two papers cited in Section 1: Zwick and Mahon ((2017); henceforth ZM) and House and Shapiro ((2008); henceforth HS).

These papers focus on policies allowing for accelerated depreciation of equipment investment that were in force in the United States during 2001–04 and 2008–11. Accelerated depreciation differs from the investment tax break we study – it brings forward depreciation deductions to the first year of an asset's life, rather than simply increasing their aggregate nominal value. This leads to additive and multiplicative adjustments to the present value of depreciation deductions. If θ represents the share of depreciation deductions accelerated into the first year, then:

Accelerated depreciation is more valuable to businesses with longer-lived capital, for whom Z is lower. Both studies use this cross-sectional heterogeneity in the effective lives of capital, and therefore the value of accelerated depreciation, for identification.

ZM use firm-level US tax data in a continuous DD set-up, as they have data to measure ZAD accurately across industries. They find a relationship between investment and the cost of capital that is significantly stronger than that found in the early literature. Their estimates indicate that accelerated depreciation raised eligible investment by 10 per cent during 2001–04 and by 17 per cent in 2008–10, relative to ineligible investment. These are significant changes, given that accelerated depreciation increased average Z by around 5 percentage points in the first period and 7 percentage points in the second period. As noted above, the investment tax break we study raised Z by between 10 and 50 percentage points.

HS build a structural model of investment, and show that under certain assumptions, the parameters governing the relationship between tax incentives and investment can be easily estimated. The assumptions they require are that the tax incentive is sufficiently temporary, and that capital is sufficiently long-lived. Under these assumptions they show that the intertemporal elasticity for investment is essentially infinite, as businesses face very large incentives to bring forward investment in long-lived capital to periods when they face lower after-tax prices. This implies that the responsiveness of investment to changes in the cost of capital (i.e. the supply elasticity) can be estimated directly from data on investment quantities. Using industry-level data on investment quantities from the first bonus depreciation period, HS estimate a supply elasticity that – once data differences are accounted for – is similar to that estimated by ZM.

HS also provide a useful discussion of earlier work that found little effect of tax incentives on investment. Goolsbee (1998) shows that investment tax incentives in the United States have a significant effect on capital prices and concludes that the supply elasticity is low. HS point out that with a very high demand elasticity, before-tax prices adjust to reflect the full cost of any subsidies, so focusing on prices is not a robust way to estimate supply elasticities. In addition, HS do not find similar price effects to Goolsbee in their analysis. HS also point out that the DD analysis performed by Cohen and Cummins (2006) used treatment and control groups that both received quite similar benefits from the policy under examination, and that this explains why they found the policy to be ineffective.

There has been less empirical analysis of the relationship between taxes and business investment in Australia. This is despite a rich history of relevant changes in taxation, including a sharp change in the depreciation schedule in 1999 (Reinhardt and Steel 2006). La Cava (2005) uses panel data to analyse the importance of financial factors on company investment, but does not examine the effect of legislative changes to depreciation schedules. AlphaBeta (2018) is a recent example that examines changes to the company tax rate for small companies that occurred in 2015. They find that affected businesses increased investment and employment.

Australia offers an interesting prism through which to examine the interaction between tax changes and business investment, given the different tax arrangements for unincorporated businesses and companies, with the latter having access to dividend imputation. A number of papers have found that dividend imputation affects the financial choices of companies in Australia. Callen, Morling and Pleban (1992) find that dividend imputation increased dividend payouts by about 20 per cent in the five years after its introduction. Twite (2001) presents evidence that businesses adjusted their capital structure away from debt after the introduction of dividend imputation.

Officer (1994) provided a user cost of capital formula applicable to a company under dividend imputation:

The difference in the user cost equations is γ ∈[0,1], which measures the value of a dollar of tax paid at the company level to a shareholder. This value reflects the fact that, under dividend imputation, tax paid at the company level is a credit for personal income tax for resident shareholders. If the value of γ is 1, indicating that the tax paid at the company level lowers shareholders' tax dollar-for-dollar, then company taxation does not appear in the user cost formula. Moreover, Z does not affect the cost of capital in this situation.

The evidence on the value of γ in Australia is mixed. Many large Australian companies raise equity from international investors, who derive no benefit from dividend imputation. As such, γ is unlikely to be 1 for these businesses. Applied evidence based on samples of listed companies finds a range of values for γ (Swan 2018). Our analysis focuses on smaller companies, who have very few non-resident shareholders. A company owned entirely by residents should have a value of γ close to 1 (this is discussed further below).