A comparison of earnings persistence in high-tech and non-high-tech firms

The pharmaceutical industry may seem incongruous with the rest of the industries in the high-tech group, other than the high cost of R&D. The pharmaceuticals are selected into the high-tech group because the extent to which accounting practices in this industry would generate unrecorded intangible assets is very high.

Improvements to software, hardware and IT processes can provide Securities brokers and dealers and other investing firms with a competitive advantage, and therefore these firms invest substantially in improving technology. Trading activities and services provided in these firms are influenced to a great extent by uncertainties in the markets. During periods of volatile markets, in uncertain regulatory or economic environment, there is a greater degree of variance between revenues and expenses in these firms.

Chang and Su (2010) recently demonstrate that R&D intensity in the electronics industry has a positive (negative) impact on sales growth when the level of such R&D intensity is below (above) the threshold value. Their evidence implies that the level of R&D intensity, relative to a threshold value, matters to the financial health of high-tech firms.

The sample selection procedure is consistent with those reported in Table 1A in Kwon and Yin (2006) and Table 1 in McGuire et al. (2012).

These firms are excluded as Cheng and Warfield (2005) and Burgstahler and Eames (2003) find that managers in these regulated industries have different motivations to manage earnings.

For the sample period, the average total assets and sales of NHT firms are approximately four times greater than those for HT firms. The average total assets (sales) of NHT firms is $3,706.88 million ($3,269.5 million), compared with $926.9 million ($844.59 million) for HT firms.

Givoly and Hayn (2000) also use assets and sales to control for inflation. In this study, since we compare accumulated nonoperating accruals between high-tech firms and low-tech firms year-by-year, the difference cannot be attributed to inflation.

TACCR_i,t = ΔCA_i,t − ΔCL_i,t − ΔCash_i,t + ΔSTD_i,t − Dep_i,t, where, for firm i at time t, ΔCA_i,t = change in current assets (item #4); ΔCL_i,t = change in current liabilities (item #5); ΔCash_i,t = change in cash and cash equivalents (item #1); ΔSTD_i,t = change in debt included in current liabilities (item #34); and Dep_i,t = depreciation and amortization expense (item #14). We also calculate total accruals as the difference between net income and operating cash flows and results are similar.

In 2006 the FASB issued FAS 123(R) requiring companies that grant stock options to expense them. Since investment projects in high-tech firms usually are long-term, and new products take years to develop, HT firms are more likely to issue options to increase the horizon of managerial action and encourage managers to take on positive net projects that improve the future prospects. In addition, because HT firms require greater cash flow availability to develop new products and services, they are likely to substitute cash compensation with stock options. High option expense may affect the barrier to entry. We do not examine the use of stock options in this paper because before 2006 there was no requirement for firms to expense stock options.

Baginski et al. (1999) and Lev (1983) link the relatively high earnings volatility of capital-intensive firms to operating leverage measured by the proportion of fixed costs to total costs.

A number of statistical programs are available to perform the principal component analysis. In this study, we use the Statistical Analysis System (SAS) to calculate the IOS index. Procedures of the principal component analysis are discussed in detail in Sharma (1996, 67–71).

In this paper, we find that (1 − Θ) = 0.914 (0.941) for HT (NHT) firms during the sample period of 1993–2005, which implies that Θ = 0.086 (0.059). Baber et al (1998, p. 176) use earnings per share data from 1974 to 1993 and find that mean value of the IMA parameter (1 − Θ) is 0.857 and hence Θ = 0.143. The difference between their results and our results possibly reflects differences in the sample period, sample size, and the data availability requirement. This study also reports the results based on an alternative time-series model, Integrated Autoregressive Process [ARI (1,1,0)].