The growth companies puzzle: can growth opportunities measures predict firm growth?

Abstract

While numerous empirical studies include proxies for growth opportunities in their analyses, there is limited evidence as to the validity of the various growth proxies used. Based on a sample of 1942 firm-years for listed UK companies over the 1990–2004 period, we assess the performance of eight growth opportunities measures. Our results show that while all the growth measures show some ability to predict growth in company sales, total assets, or equity, there are substantial differences between the various models. In particular, Tobin's Q performs poorly while dividend-based measures generally perform best. However, none of the measures has any success in predicting earnings per share growth, even when controlling for mean reversion and other time-series patterns in earnings. We term this the ‘growth companies puzzle’. Growth companies do grow, but they do not grow in the key dimension (earnings) theory predicts. Whether the failure of ‘growth companies’ to deliver superior earnings growth is attributable to increased competition, poor investments, or behavioural biases, it is still a puzzle why growth companies on average fail to deliver superior earnings growth.

Keywords:

• growth opportunities
• growth proxies
• firm growth

JEL Classification :

• G31

Acknowledgements

We are grateful to Chris Adcock (the editor), Seth Armitage, Henk Berkman, Sheng-Hung Chen, Alan Goodacre, Nobuyuki Iwai, David Power, Johannes Raaballe, Richard Taffler, Isaac Tabner, Chris Veld, and two anonymous referees for suggestions and remarks on previous versions of this paper. We also acknowledge useful comments received at presentations at the Multinational Finance Society Annual Conference (Orlando, 2008), the British Accounting Association Scottish Conference (Glasgow, 2008), the British Accounting Association Annual Conference (Dundee, 2009), the European Financial Management Association Annual Conference (Milan, 2009), as well as research seminars at Stirling University and the University of Glasgow.

Notes

The agency theory suggests investments in negative NPV projects may be undertaken by managers if there is a separation between ownership and control (e.g. Jensen 1986; Carroll and Griffith 2001; Andrikopoulos 2009).

Notable studies using market-to-book ratio measures as proxies for the level of growth opportunities include Collins and Kothari (1989), Chung and Charoenwong (1991), Smith and Watts (1992), Gaver and Gaver (1993), Kallapur and Trombley (1999), Jacquier, Titman, and Yalcin (2001), Burton (2003), and Adam and Goyal (2008).

See, e.g. Kester (1984), Chung and Charoenwong (1991), Penman (1996), Jacquier, Titman, and Yalcin (2001), and Kallapur and Trombley (1999).

See, e.g. Rozeff (1982), Gaver and Gaver (1993), Smith and Watts (1992), Kallapur and Trombley (1999), and Jacquier, Titman, and Yalcin (2001).

We have not included a number of other variables that make more fleeting appearances in the literature, such as gearing, R&D, and share price volatility.

While the model was included in the first edition (1981) of the Brealey and Myers textbook, the model was more fully discussed and given more prominence from the fourth edition (1991) onwards.

K_s is estimated using the CAPM: , where K _f refers to the risk-free interest rate and K _m to the return in the stock market index. In the calculations that follow, we use β estimates obtained from the London Business School Risk Measurement Service, edited by Dimson and Marsh. We follow Hirst, Danbolt, and Jones (2008) and assume an equity risk premium of 6%, which is towards the middle of the estimates put forward in the literature (Dimson, Marsh and Staunton 2003). However, as argued by Fama and French (1997, 178), ‘Estimates of the cost of equity are distressingly imprecise. Standard errors of more than 3.0% per year are typical when we use the CAPM or the three-factor model to estimate industry CE's [cost of equity]’. The costs of capital used in our estimates are similarly subject to considerable measurement error. As the cost of capital will impact on the estimated value of P _a and P _g (see Danbolt, Hirst, and Jones 2002), the value of growth opportunities reported should be taken as estimates. However, changing the cost of capital is unlikely to have a significant impact on the relative importance in the value of growth opportunities to the various sample firms, or to the correlation and regression results reported below.

We use the average of the consensus I/B/E/S earnings forecast, at the financial year end, of current year earnings (F1MN) and earnings for the following year (F2MN), obtained from Datastream.

K _sr is estimated using a real risk-free rate in the CAPM, as follows: . We use the yield on index-linked gilts as the real risk-free rate.

The latest edition of the Brealey and Myers’ text (Brealey, Myers, and Allen 2006) includes an example in which the calculated value of growth opportunities is negative. They note the difficulty of interpreting this result.

The KBM model estimate the value of assets in place based on K _s (using the overall equity beta, β_s) rather than the cost of capital for assets in place, K _a (using the beta for assets in place, β_a).

D ₀ refers to the dividend for the year, g to the (real) constant rate of growth, and E to the book value of equity per share. Note that the HDJ model discounts using the real cost of capital (K _ar), as in KBM_real.

In their study of the dividend behaviour of UK-listed companies over the 1990s, Renneboog and Trojanowski (2005) found that 85% of the firms pay dividends, with dividend yields averaging 3.1%.

MV refers to the market value and BV to the book value. BV debt is calculated as the sum of book values of loans and short-term debt.

We base the analysis on EP rather than the inverse PE ratio to reduce the influence of outliers when EPS is small.

In the EP model, we use EPS₀ and restrict the analysis to where EPS₀ is positive.

The subsequent analysis of future earnings growth is based on EPS data for the period from January 1989 to December 2007.

There is overall relatively little difference in the correlations whether we base the analysis on Pearson's or Spearman's (rank) correlations. We therefore simply report and discuss the Pearson's correlations.

Data from Thomson ONE Banker database suggest that 30% of the acquisitions by listed UK companies during our sample period involved payment in shares, and – as the share-financed transactions tended to be larger – ordinary shares accounted for almost 50% of the total consideration paid in these acquisitions.

In addition to the 3.3% of the initial sample with negative book values (Table 1), a large proportion of our sample firms have small book values. While Nissim and Ziv (2001) find only 0.6% of their sample of US dividend-paying companies to have book values of less than 10% of their total assets, the comparable figure for our final sample is 14.3%.

In our analysis of equity growth, we include all firm-years with positive book values. We have also undertaken the analysis (i) with the sample restricted to cases where book equity accounts for at least 10% of total assets and (ii) resetting book to 10% of total assets where book values are below this threshold. The results are overall very similar and our conclusions unaffected.

Earnings growth measures are only calculated if data are available for each firm-year in that earnings measure. Thus, for EPS Grow₊₂₊₄, EPS data must be available for years −1, 1, 2, 3, and 4. However, in order to reduce survivorship bias, we do not require data to be available for all growth measures. In the calculations that follow, we use the EPS measure WC05201 (obtained from Datastream), which include negative earnings. (Note the Datastream variable ‘EPS’ gives a value of zero for loss-making firms and is as such not suitable for our analyses.)

As a robustness test, we have also undertaken the analysis based on excess earnings growth, where we control for general movements in average earnings over our sample period by subtracting the average earnings growth for UK companies during the period of analysis from the measured level of firm growth. The results regarding the relationship between growth opportunities and excess earnings growth are very similar to those for earnings growth. For brevity, we do therefore not report the full excess earnings results.

The results for realised excess earnings growth are very similar to those reported above for earnings growth. Over the short period, Q is significant at the 10% level. However, no other purported growth measure is significant at any level. Over the medium term, only HDJ and EVE are significant (at the 5% level), but the relationship has the wrong sign. For our long period, HDJ and EVE still show up as significant but with the wrong sign, while only DP manages to be both significant (at the 5% level) and to have the right sign. Generally, the growth opportunity measures perform no better as a predictor of excess earnings growth than they did for unadjusted earnings growth.

Analysis of the correlation matrix reveals that ROE₋₁ is significantly negatively related to future earnings growth on a univariate basis, but the variable loses significance in the regression model once we also include Δ EPS. (The correlation between ROE₋₁ and Δ EPS is −0.276, significant at the 1% level.) All other correlations are consistent with the regression coefficients reported in Table 7. None of the correlations between the independent variables in Table 7 exceed 0.4, and similarly none of the variance inflation factors (VIF scores) for Table 7 exceeds 1.4. This suggests that there is no significant issue of multicollinearity in the analysis. For brevity, we do not report the full correlation matrix, but it is available from the authors upon request.

In an empirical study of seasoned equity offers in the UK, Andrikopoulos (2009) find that issuers generate high turnover at the expense of profitability for the first 2 years following the equity offers, and that sales revenue also slows down approximately 3 years after the issue. Andrikopoulos ‘…suggests that the deterioration in the operating performance is caused by the potential existence of managerial hubris and “empire-building” biases on behalf of the issuers’ (p. 190).

We acknowledge, however, that with information asymmetry, it may be difficult for investors to observe the link between marginal investment and marginal earnings.