Is UK business investment unusually weak?

Barrell, Ray ; Riley, Rebecca

In the 15 years to 2000 UK business investment rose quite substantially in real terms. Having shown little deviation from an average of 7 1/4 per cent over the period 1965-85, the real share of business investment in GDP measured 10 3/4 per cent in 2000 (see figure 1). One of the factors contributing to this increase was the sharp fall in the relative price of investment goods, the scale of which is illustrated by the altogether different trends in real and nominal shares of business investment in GDP. (1)

[FIGURE 1 OMITTED]

As discussed in the Commentary section of this Review (pages 4-9), recent concern is that investment is unusually weak, both in the UK and elsewhere. Certainly in the UK, when measured in nominal terms, business investment has remained at a historic low relative to the rest of the economy since 2002. At the same, measured in real terms, the share of business investment in GDP has fallen back a bit from its peak at the turn of the century. This is despite further declines in the price of investment goods to other goods and the relatively low cost of finance, although as we discuss below the latter is more apparent on some measures than others.

Here we assess the recent performance of UK business investment within a simple aggregate framework, similar to that which underlies the forecast and to analysis elsewhere (e.g. Bakhshi and Thompson (2002); Ellis and Price (2004)).

Standard economic theory suggests that the demand for capital as an input into the production process is determined by the real user cost of capital, the parameters of the production technology and the mark-up over unit costs. Assuming a CES technology with elasticity of substitution [sigma], a constant mark-up over unit costs and zero capital augmenting technical progress, and making use of the result that in long-run equilibrium capital and investment grow at the same rate (as in Bean, 1981) the long-run relationship between investment, output and the user cost of capital is written as

log [IB.sub.t+1] = cons + log [Y.sub.t] - [sigma]log[USERC.sub.t] (1)

where IB is investment (volumes), Y is output (volumes) and USER C is the real user cost of capital.

We estimate an error-correction equation for business investment around the long run in (1), approximating business sector output with economy wide output and including a measure of industry capacity utilisation (2) to help explain short-run deviations from the long run as

[DELTA] log[IB.sub.t] = cons + [[alpha].sub.ECM] log[(IB/Y).sub.t-1] + [[alpha].sub.USERC] log[USERC.sub.t-1] + [[alpha].sub.CU][CU.sub.t] + dynamics (2)

where [sigma] = [[alpha].sub.USERC]/[[alpha].sub.ECM]. (3)

We use a Hall-Jorgensen real user cost of capital, where the real cost of finance is a standard weighted cost of capital measure. (4) The importance of taking into account the cost of both debt and equity finance for the current purposes is illustrated in figure 2. There we show estimates of the real user cost of capital assuming that all finance is either debt or equity based. The user cost of capital that takes into account the real cost of debt finance looks particularly low at present, as real long-term interest rates have fallen quite far in comparison to the past 25 years. In comparison, the user cost of capital that takes into account the real cost of equity finance looks less depressed, at least relative to the latter part of the 1990s, because the return on equity has not fallen in the same way.

[FIGURE 2 OMITTED]

The relationship in (2) for UK business investment is reported in table 1, based on data for the past 30 years. The equation gives an elasticity of substitution between labour and capital of 0.48, consistent with the results of Oulton and Young (1996) and Ellis and Price (2004) and with the estimated elasticity of substitution from the labour demand curve used in NiGEM. (5) The equation diagnostics suggest that the equation is well determined, although the Chow test for the stability of the regression coefficients over the first and second halves of the sample indicates that the equation is not as stable as one might prefer, (6) and the significance of the error correction coefficient indicates cointegration.

Conducting a series of simple tests we evaluate whether or not business investment in recent years should be regarded as particularly different from the past. Augmenting the equation with a dummy variable equal to one for the period 2003q1-2005q4 and zero otherwise, growth in business investment appears to be marginally weaker in the past three years, but not significantly so (dummy coefficient estimate -0.00639; t-statistic 0.72). Shortening the sample to the end of 2002, we can generate a forecast of growth in business investment. The forecast errors are generally negative (see figure 3) showing that actual growth in business investment over the period 2003-5 was lower than the simple model would predict. But, a predictive failure test does not suggest that business investment has been significantly different from that which the equation would predict ([chi square](12) = 8.81 (0.72)). In contrast, a test for structural stability over the period 2003-5 indicates that there may be some prediction failure ([chi square](4) = 8.20 (0.084)), at least at the 10 per cent level. The other coefficients of the equation are largely unchanged in these exercises.

[FIGURE 3 OMITTED]

Thus, based on the analysis of aggregate business investment here it is difficult to say that growth in business investment volumes has been particularly weak in the UK in recent years. Of course analysis of aggregate business investment may be associated with a number of pitfalls (as emphasized in Bakhshi et al., 2003). With such caveats in mind, the message from aggregate analysis is that the evidence for significant underinvestment is not strong. If the present situation of slight underinvestment were to persist, the conclusions reached here may be different.

REFERENCES

Bakhshi, H., Oulton, N. and Thompson, J. (2003), 'Modelling investment when relative prices are trending: theory and evidence for the United Kingdom', Bank of England Working Paper No. 189.

Bakhshi, H. and Thompson, J. (2002), 'Explaining trends in UK business investment', Bank of England Quarterly Bulletin, Spring, pp. 33-41.

Bean, C. (1981), 'An econometric model of manufacturing investment in the UK', Economic Journal, 91, pp. 106-21.

Ellis, C. and Price, S. (2004), 'UK business investment and the user cost of capital', Manchester School, 72 S1, pp. 72-93.

Oulton, N. and Young, G. (1996), 'How high is the social rate of return to fixed investment?', Oxford Review of Economic Policy, 12, pp. 48-69.

NOTES

(1) See Bakhshi and Thompson (2002) for a discussion of the different factors that may have contributed to the rise in UK business investment in the latter decades of the 1900s.

(2) CBI Quarterly Industrial Trends Survey, Table 1, question 4.

(3) We also include dummy variables for the first two quarters of 1985 to account for the sharp increase in investment that followed changes in tax allowances.

(4) [USERC.sub.t] = [RP.sub.t]([r.sub.t] + [[delta].sub.t] - [DELTA] log [RP.sup.e.sub.t]) I/(I-[CTAXR.sub.t]) where RP is the ratio of the National Accounts business investment and GDP deflators, [r.sub.t] = 0.25[lrr.sub.t] + 0.75[ep.sub.t] where lrr is the real rate of interest on 10-year government bonds and ep is the ratio of gross operating surplus to the value of net equity for private non-financial companies, [delta] is the depreciation rate for the whole economy capital stock, and ctaxr is the effective corporation tax rate. Expected relative price movements are set to average outcomes for the past two years as in Bakhshi et al. (2003).

(5) National Institute Global Econometric Model.

(6) Judging by the cumulative sum of recursive residuals there are issues of parameter instability from the mid-1990s.

Ray Barrell and Rebecca Riley *

* National Institute of Economic and Social Research. Thanks to Martin Weale for helpful comment.

Table 1. UK business investment (OLS estimates)

Parameter Estimate t-statistic

cons -0.29167 (4.80) [[bar.R].sup.2] 0.3769
[alpha]CM -0.08331 (3.50) SE 0.0259
[alpha]ERC -0.04036 (2.99)
[alpha]CU 0.21907 (5.01) Serial Corr. [chi square](4)
 = 5.40 (0.249)
 RESET [chi square](1)
 = 0.46 (0.498)
 Norm [chi square](2)
 = 0.33 (0.850)
Sample 1976q1-2005q4 Hetero [chi square](1)
 = 0.95 (0.329)
 Chow [chi square](4)
 = 8.48 (0.075)