Fiscal incentives for research and development

Abstract

It is often argued that since the social return to R&D exceeds the private return, the government should provide incentives for R&D expenditure. This article considers the issue of the impact of such incentives on the fiscal position of the government, using a simple comparative static model. In particular, it is argued that it is possible that the social return from R&D might be sufficient to allow R&D incentives to more than pay for themselves. On the basis of the international evidence, the model is calibrated to examine what values of the key parameters are required in order for this conclusion to hold.

Acknowledgement

The authors acknowledge the academic and research input of Economic Research Southern Africa.

Notes

¹ The classic exposition is Arrow (1962). For a more recent discussion, see Hall (1996). Some supporting empirical findings are provided by Frantzen (1998), Meliciani (2000) and Smith et al. (2004); Atella and Quinteri (2001) provides countervailing evidence.

² See for instance the discussion in Kwack and Sun (2005), though see the potential allocative efficiency costs that may be associated with R&D activity noted in Yoon (2004).

³ We employ the inverse of the B-index, which is computed as B = ATC/(1 − t), where ATC is the after tax cost of a $1 expenditure on (say) R&D, and t is the tax rate on corporate income. Thus a lower score on the B-index indicates a stronger tax incentive. To avoid the counterintuitive scaling, we invert the index. See the discussion in Warda (1996).

⁴ Spillovers associated with learning and innovation of course have a long history in economics. See for instance the seminal Arrow (1962) and Romer (1986) contributions, and the more recent empirical evidence in Dietzenbacher (2000) and Diao et al. (2006).

⁵ For a paper which explicitly addresses the international transfer of knowledge in a tax incentive model, see Griffith et al. (2001).

⁶ Input prices are measured in units of the output good.

⁷ All types of expenditure can be fully deducted once for purposes of calculating taxable profits. Accordingly, a value of δ = 0.5 must be interpreted as meaning that a firm can claim one and a half times the amount of its R&D expenditure as a tax deduction. Thus the form of the tax incentive that we are considering is one in which the government allows a deduction against the firm's taxable income, although in practice there are many different ways in which governments have structured tax incentives. Note also that we assume that R&D funded by government grant cannot also be claimed by the firm as a tax deduction.

⁸ Generally we set 0 < λ < 1. This presumes that increasing returns, if present, are not strong enough to generate sufficient additional tax revenue to render more than proportional subsidy support on the part of the fiscal authority rational.

⁹ Note that this causes the terms containing the effect of labour and capital to disappear from the tax equation. This is because at the profit-maximizing level, the marginal products of these factors of production are set equal to their prices. Thus, when there is any marginal change in the employment of these factors, the effect on tax revenue due to the resulting change in output is exactly offset by the change in the amount of expenditure that can be deducted from taxable profits.

¹⁰ Note that in deriving these expressions, we have used the assumption made in Equation 1 that the cross-partial derivatives Y _{R i L i} and Y _{R i K i} are zero.

¹¹ Since we have shown in Equation 10 that is a positive function of , we will limit our proofs below to the case of changes in the tax allowance, since the idential conditions will apply to changes in the level of the grant.

¹² Again, since is a positive function of , in the proofs we will only consider changes in the rate of tax allowance.

¹³ On the Israeli case, see the discussion in Trajtenberg (2001).

¹⁴ In terms of grants, from 1995–1999 a total of 6593 projects were approved, for a total of US $ 1919 million.

¹⁵ The evidence also suggests that Israel has made the transition evident in worldwide trends in patent registration from Mechanical and Chemical patents, to Medical and Computer & Communications Technology more rapidly than the US.

¹⁶ In fairness it must be pointed out that Malaysia's technology policy started in the mid-1980s, while the foundations of Singapore's were laid in the 1960s with its aggressive investment in human capital. Thus the differential effectiveness may simply be a function of lag effects – technology policy is inherently a long-term growth policy which takes time to realize.

¹⁷ There is discussion as to whether the reason for the failure is demand side (the result of tastes and abilities), supply side (number of student places on offer), or the impact of affirmative action for Bumiputeras, but evidence is not conclusive. See the discussion in Mani (2000) and Lall (1999).

¹⁸ See for instance the evidence on the US experience, contained in Baily and Lawrence (1987, 1992), Tillinger (1991), Swenson (1992), Berger (1993), Hall (1993), Hall and van Reenen (1993, 1999), Hines (1993) and Shah (1995).