The determinants of the choice of accounting for software development costs in the US software industry

The determinants of the choice of accounting

for software development costs in the US

software industry

Faculty:

Faculty of Economics and Business, University of Amsterdam

Program:

MSc Accountancy and Control – Accountancy track

Student:

Bryan Kelder

Student number:

10180524

Supervisor:

Vincent O’Connell

Date:

17

th

of August, 2015

Academic year:

2014-2015

2

Statement of Originality

This document is written by student Bryan Kelder who declares to take full

responsibility for the contents of this document. I declare that the text and the work presented

in this document is original and that no sources other than those mentioned in the text and its

references have been used in creating it. The Faculty of Economics and Business is

responsible solely for the supervision of completion of the work, not for the contents.

Abstract

This paper investigates what the determinants are of capitalizing versus expensing

software development costs in the US computer programming and prepackaged software

industry. Using earnings volatility, earning sign, size, leverage, R&D intensity, and

market-to-book value as independent variables, and the choice to capitalize as dependent variable, I find

that earnings volatility, earnings sign, size, and R&D intensity are associated with the

decision to capitalize software development costs (respectively, negatively, positively,

positively, and negatively associated). It can be concluded that firms in the US software

industry are conservative in regard to capitalizing software development costs. This is based

on the results that firm that capitalize are generally larger, have lower earnings volatility, and

have positive earnings. This study contributes to the existing literature in three ways. Firstly,

my research offers insight on the determinants for the decision to capitalize versus expense

software development costs. Policymakers can use the results of this study when deciding

upon issues pertaining to R&D accounting in the US. Furthermore, the FASB and IASB have

announced that they want to converge their accounting rules, so one set of accounting rules of

high quality can be made. This study offers insight in the way US firms handle the

capitalizing of costs, which is important as IFRS uses cost capitalization in more instances

than US GAAP. Additionally, this study contributes to the earnings management literature.

However, while earnings management was not directly studied, it was found that the leverage

variable did not influence the choice to capitalize software development costs. Finally, this

study contributes to the existing literature on determinants of accounting choices. However,

the results must be interpreted carefully, as the sample contains far less capitalizers than

expensers. Furthermore, capitalizing firms also expense R&D costs, and finally, it is assumed

that all R&D expenses are for the benefit of software development projects. Moreover,

because the sample is from the US computer programming and prepackaged software

industry, one needs to apply caution when trying to generalize these findings to other

industries in the US.

3

Acknowledgements

Here I present to you the product of four years of university education at the University of

Amsterdam. In the years that I have studied there, I have found a wealth of knowledge which

I happily accepted. Through all this time, my parents gave me a warm home and the

opportunity to educate myself through an university education. I want to thank for parents for

giving me the chance that they never had when they were young; making me aware of the

difficulties that they encountered in life, and how I could benefit from working hard for my

education. Furthermore, I want to thank my sister for the virtue of being there. Moreover, I

want to thank the family that helped me during difficulties that I encountered while having to

work on my thesis, and I could not. Last, but not least, I want to thank my supervisor for

constantly helping me with my thesis, even when he did not have to do so. Without the help

of my supervisor, I would not have finished my thesis in time, so thank you for that.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4

Table of Contents

1. Introduction ... 5

2. Accounting for R&D in the US ... 7

3. Determinants of the choice to capitalize R&D versus expense R&D ... 10

4. Hypothesis development and research design ... 13

5. Data and sample description ... 21

6. Univariate and multivariate results ... 23

7. Conclusion ... 26

5

1. Introduction

In the United States of America, there are two possible treatments to account for

research and development programs. The main treatment used in the US is the rule stated in

the Statement of Financial Accounting Standards No. 2, or SFAS No. 2. It states that a direct

relationship between research and development costs and particular future revenue has

generally not been observed, even with the benefit of hindsight (Financial Accounting

Standards Board, 1974). Moreover, it states that it assumes this relationship, because they

base this on three empirical studies done, and they found that there was no significant

correlation between R&D expenditures and increased future benefits. SFAS No. two contains

values that supporters of the expensing alternative would suggest to be of high value. They

state that expensing research and development costs would lead to more comparability and

consistency. Moreover, they state that financial data would be more objective (Markarian et

al., 2008). They also state that there would be less opportunities to manipulate earnings, as

managers have incentives to capitalize projects that will not succeed.

However, there is an exception to the rule, and these rules are described in SFAS No.

86, or accounting for the costs of computer software to be sold, leased, or otherwise marketed.

It states that costs incurred internally in creating a computer software product shall be charged

to expense when incurred as R&D until technological feasibility has been established for the

product. So that means that this financial accounting standards does not assume that there is

no relationship between research and development costs and a particular future revenue,

because firms that use SFAS No. 86 can capitalize their software costs on the balance sheet,

and amortize them instead of directly expensing their R&D costs. The capitalization of

research and development expenses is motivated by three argument that support the use of

capitalization over expensing, which are the positive influence on market value, the positive

influence on earnings informativeness, and the positive influence on future income.

The last 40 years have witnessed an immense growth of R&D investment in the US

and other countries and the emergence of new R&D-intensive industries, therefore these rules

have become more important these last years (Cohen, Dey, & Lys, 2008). Moreover, research

on the determinants to capitalize versus expense R&D costs is scarce. There is a segregation

between firms that use SFAS No. 2, and SFAS No. 86. Therefore, an examination of the

determinants of the decision to capitalize or to expense software development costs provides

information about the reasons underlying the firm’s decision. These results may have

6

Therefore, the objective of this paper is to address the question of what the

determinants of expensing versus capitalizing software development costs in the US are.

Based on prior research, it is hypothesized that the decision to capitalize versus

expense software development costs in the US is influenced by four variables: earnings

volatility, earnings sign, and firm size, and R&D intensity. Using a final sample of 5448

unique firm-year observations in the period 1998-2006 from the US programming and

prepackaged software industry, this study finds that the decision to capitalize is negatively

associated to earnings volatility (firms with lower earnings volatility are more likely to

capitalize), while it was hypothesized that earnings volatility would be positively related to

the choice to capitalize. Furthermore, this study also hypothesizes that earnings sign is

negatively related to the choice to capitalize software development costs. However, this study

finds that earnings sign is positively related to the choice to capitalize software development

costs (firms with positive earnings are more likely to capitalize). Additionally, this study

hypothesizes that firm size is positively related to the choice to capitalize software

development costs. It is found that this prediction is true based on the results (firms of a

greater size are more likely to capitalize). Finally, in this study it is hypothesized that R&D

intensity would be positively related to the choice to capitalize, however the R&D intensity

variable was found to be negatively associated with the decision to capitalize software

development costs (firms with a lower R&D intensity are more likely to capitalize).

Overall, this study finds evidence that US firms in the American software industry

generally find it not advantageous to capitalize their software development expenses. Based

on the results, it seems that these firms are very conservative when it comes to capitalizing

these costs. This is based on the results that firm that capitalize are generally larger, have

lower earnings volatility, and have positive earnings.

This study contributes to the existing literature in three ways. Firstly, my research

offers insight on the determinants for the decision to capitalize versus expense software

development costs. Policymakers can use the results of this study when deciding upon issues

pertaining to R&D accounting in the US. Furthermore, the FASB and IASB have announced

that they want to converge their accounting rules, so one set of accounting rules of high

quality can be made. This study offers insight in the way US firms handle the capitalizing of

costs, which is important as IFRS uses cost capitalization in more instances than US GAAP.

Additionally, this study contributes to the earnings management literature. However, while

earnings management was not directly studied, it was found that the leverage variable did not

influence the choice to capitalize software development costs.

7

This study proceeds as follows. Section 2 introduces accounting for R&D in the US.

Section 3 discusses the previous literature and the relation which this study has with it.

Section 4 presents the development of hypotheses and research design, and this is followed by

the description of the data and the sample that is used in this study in section 5. Furthermore,

section 6 presents the results of the research and section 7 concludes the study.

2. Accounting for R&D in the US

Accounting for research and development expenditures has little flexibility when it

comes to capitalization of R&D costs. There are two major financial accounting standards

under US general acknowledges accounting principles (hereafter called US GAAP). These

two are called the statement of financial accounting standards No. 2 (SFAS No. 2) which

addresses the accounting for research and development costs, and the statement of financial

accounting standard No. 86 (SFAS No. 86) which addresses the accounting for the costs of

computer software to be sold, leased, or otherwise marketed.

According to SFAS No. 2 there are two stages in research and development:

1. “Research is planned search or critical investigation aimed at discovery of new

knowledge with the hope that such knowledge will be useful in developing a new

product or service (hereinafter "product") or a new process or technique (hereinafter

"process") or in bringing about a significant improvement to an existing product or

process.

2. Development is the translation of research findings or other knowledge into a plan or

design for a new product or process or for a significant improvement to an existing

product or process whether intended for sale or use. It includes the conceptual

formulation, design, and testing of product alternatives, construction of prototypes,

and operation of pilot plants. It does not include routine or periodic alterations to

existing products, production lines, manufacturing processes, and other on-going

operations even though those alterations may represent improvements and it does not

include market research or market testing activities” (Financial Accounting Standards

Board, 1974).

All research and development costs are to be expensed as incurred. So one cannot capitalize

these costs on the balance sheet. The statement does give an explanation for the fact that costs

8

can only be expensed as incurred. The statement states that there is a high degree of

uncertainty about the benefits that research and development projects will provide, they

substantiate this argument with a study where it was found that less than two percent of

product ideas and less than fifteen percent of product development projects were brought to

the market. Additionally, the statement states that there is a lack of a causal relationship

between the expenditures made in the research and development period and the benefits

obtained after the research and development period. They substantiate this argument by

stating that three studies failed to measure a significant causation between R&D expenditures

and increased future benefits. Finally, it is stated that based on anecdotal evidence that

capitalization would not communicate additional useful information to investors.

However, under SFAS No. 86, a part of SFAS No. 2 is used. It is specified that all

costs incurred to establish the technological feasibility of a computer software product to be

sold, leased, or otherwise marketed are research and development costs. So, these costs are

required to be expensed as incurred just as R&D expenditures under SFAS No. 2. The

difference between the two standards comes from the fact that R&D costs can be capitalized

when the technological feasibility of a computer software product is established. According to

SFAS No. 86, there is evidence that the technological feasibility is established if one of the

following requirements are fulfilled:

1. “If the process of creating the computer software product includes a detail program

design:

(a) The product design and the detail program design have been completed, and the

enterprise has established that the necessary skills, hardware, and software technology

are available to the enterprise to produce the product.

(b) The completeness of the detail program design and its consistency with the product

design have been confirmed by documenting and tracing the detail program design to

product specifications.

(c) The detail program design has been reviewed for high-risk development issues (for

example, novel, unique, unproven functions and features or technological

innovations), and any uncertainties related to identified high-risk development issues

have been resolved through coding and testing.

If the process of creating the computer software product does not include a detail program

design with the features identified under 1. above:

9

2. (a) A product design and a working model of the software product have been

completed.

(b) The completeness of the working model and its consistency with the product

design have been confirmed by testing” (Financial accounting standards board, 1985).

Moreover, when it comes to amortizing the capitalized development costs, it is determined

that the amortization shall be the greater of the amount calculated using a ratio that presently

gross revenues for a product bear to the total of current and anticipated future gross revenues

for that particular product or using a straight line method over the remaining estimated

economic life of the product. However, it is also stated that amortization will only commence

if the product becomes available for the general public.

As seen above, the requirements are highly subjective, and require a lot of

documentation to be written up by the firm wanting to capitalize its software development

costs. Moreover, a manager of a firm could capitalize software development costs temporarily

of products that have failed given the subjectivity and the fact that amortization will start if

the product becomes available for the general public. This could potentially mean that

capitalizing software development costs could be costly because of the recordkeeping costs

involved. Moreover, there is room for earnings management via accruals in the standard.

Furthermore, the statement does not provide an explanation why the FASB chose to deviate

from the method used in SFAS No. 2, where the FASB does not condone capitalization of

R&D costs and gives reasons why which could also apply to software development costs.

However, it is evident that the rules laid down in SFAS No. 86 have some similarity

with the rules for R&D capitalization laid down in IAS 38 of the International Financial

Reporting Standards (IFRS). Moreover, the FASB and International Accounting Standards

Board (IASB) announced that they are committing to the convergence of US and international

accounting standards. Therefore, it is interesting to research what the determinants are for US

firms in the software industry which influence the choice whether to capitalize software

development costs versus expense software development costs. This could have implications

on the debate how the rules should be converged to create better standards for the US.

10

3. Determinants of the choice to capitalize R&D versus expense R&D

This study is most closely related to the study done by Oswald (2002) who has done

research on the determinants and value relevance implications of accounting method choice

for development expenditures for firms with R&D programs in the United Kingdom (UK). He

selected a sample of 3229 UK firm-years over a period of 1996-2004, and identifies that there

are seven variables which are significantly associated with the choice to capitalize R&D

expenditures versus expense. He found that those variables were: earnings variability,

earnings sign, firm size, R&D intensity, leverage, steady-state status of the firm’s R&D

program, and R&D program success. My study differs from Oswald’s in that I examine a

different accounting rule, a different accounting jurisdiction, and this may have implications

on the debate for the convergence of US GAAP and IFRS. It is often poised that countries that

fall into the same category often display the same behavior. Literature tends to categorize

countries into two categories which are: common law countries and code law countries.

Common law countries imply that there is a market-oriented country, where there is a

“shareholder” model which gives exclusive corporate rights to shareholders. Code law

countries imply that there is stakeholder-centric view. The different accounting jurisdiction is

important, because it cannot be assumed that firms in common-law countries such as the UK

and the US will exhibit the same behavior when determining the choice to capitalize or

expense research and development costs. To my knowledge, there are no other studies that

have attempted the same thing. My study also differs in the fact that Oswald uses two

variables that were found to be associated with the choice to capitalize research and which are

state of the firm’s R&D and the R&D program success. The first variable is the

steady-state status of the R&D of a firm, where it is expected that firms who only do R&D

expenditures to maintain the level of the R&D asset are more likely to expense their R&D.

This is hypothesized because for firms that are in this status, the amount that is capitalized on

the balance sheet as development expenditures will be equal to the amount that would be

amortized from the research and development asset. This implies that for these companies that

there is a steady-state, so a state where it does not matter whether the firm expenses research

and development costs, or whether it capitalizes its research and development costs. However,

it is given that a firm is subjected to costs of recordkeeping if a firm capitalizes its research

and development costs on the balance sheet, therefore it would be more beneficial for the firm

to expense its research and development costs. He finds that firms who are not in a steady

state prefer to capitalize their research and development costs, therefore substantiating his

argument. Finally, the author uses the R&D program success where he creates a value using

11

the difference between market value of equity minus book value of equity, divided by R&D

expenditures. He asserts that more successful firms have a higher value of this number. He

assumes that firms that are successful can communicate this to the market by capitalizing their

R&D expenditures. It seems that the calculation of R&D program success could be influenced

by noise, as the difference between market value of equity and book value of equity is not

solely determined by the research and development expenditures done by firms.

Additionally, a study done by Markarian, Pozza, & Prencipe (2008) has some

similarity to this study. They state that firms use the flexibility to capitalize their R&D costs

to manage earnings. Based on a sample of Italian listed firms, they examine whether

companies’ choices to capitalize R&D costs are affectuated by motives to manage earnings.

They do find that earnings-smoothing is associated with the choice to capitalize R&D costs.

However, they do not find that a firm’s level of debt financing is associated with capitalizing

R&D costs. The study that they conduct can generalize its findings to other European

countries which use IFRS, as the Italian standards for R&D accounting are almost identical to

the IFRS. This study is similar to the one of Markarian et al. (2008) in that some of the

proxies that they use for investigating the decision to capitalize R&D costs, are also used by

this study. It is different in that they only concentrate on earnings management. This study

concentrates on a wider variety of variables that are supposed to have an influence in

determining the choice to capitalize or expense R&D costs. Furthermore, there are several

studies done in which the decision to capitalize R&D costs and the decision to manage

earnings is researched, while researching the wider variety of determinants for the choice to

capitalize R&D costs is not done in a wider variety.

Additonally, it was found by that early-adopters of SFAS No. 86 were generally

smaller firms with auditors who expressed support for SFAS No. 86 while it was in the

exposure draft stage (Trombley, 1989). This study is similar in that size is a variable which is

used as a potential determinant of capitalizing software development costs. However, it

differs in that this study does not try to determine which the characteristics of early-adopters

were, but it tries to focus on the characteristics of existing incumbent members of the software

industy. This is more informative, as these incumbent members can be seen as the final result

of the SFAS No. 86 instead of a probable temporary result created by auditors who expressed

support for this standard.

Furthermore, research and development accounting has always been a controversial

accounting issue. Based on the previous literature, researchers have found benefits of

12

capitalization of research and development costs, and they have found disadvantages of

capitalizing research and development costs.

The first benefit that supporters of capitalization of research and development costs

suggest is that R&D is an asset that influences future income. This is firstly researched by

Ballester, Garcia-Ayuso, & Livnat (2003) who find the existance of R&D assets positively

influences future profitability. Moreover, a study done by Sougiannis (1994) examines

whether reported accounting income reflects benefits from past R&D expenditures. The

author uses two equations to calculate a R&D asset, one equation for earnings, and one

equation for valuation. The valuation equation determines how accounting and R&D values

are valued in the market. The two outcomes are then used to estimate the investment value of

the research and development expenditures. The results indicate that earnings which are

reported do benefit from research and development expenditures. He finds that, generally, a

increase of one dollar in R&D would lead to a two dollar increase in profits.

The second benefit that supporters of research and development cost capitalization

suggest is that research and development costs are found to be postively related to market

value. The study done by Sougiannis (1994) as described above finds evidence of this.

Furthermore, Hirschey & Weygandt (1985) also find evidence in their study where they

examine whether advertising and R&D expenditures have a positive effect on the market

value of the firm. Using a sample of 390 firms from 20 product groups, they find that both

advertising and R&D expenditures have definitive influences on the market value of a firm

that will persist over time.

Finally, there are studies that focus on research and development accounting and the

communication of value relevant information to investors. These studies generally find that

there is support for the communication of value relevant information to investors and the

capitalization of research and development expenditures (Aboody & Lev, 1998). While one of

the general concerns of SFAS No. 2 is that capitalization of research and development costs

would lead to incentives for earnings management, it is found that even when earnings

management is persistent, the relevance of the information communicated to the market

would still be more beneficial (Healy, Myers, & Howe, 2002).

However, there are also researchers that state that there are disadvantages to

capitalizing research and development costs. They state that expensing research and

development costs would lead to more comparability and consistency. Moreover, they state

that financial data would be more objective (Markarian et al., 2008). They also state that there

would be less opportunities to manipulate earnings, as managers have incentives to capitalize

13

projects that are bound to fail. When examining SFAS No. 86, it could indeed be made out

that there was flexibility to capitalize bad projects as seen in the section before (Aboody &

Lev, 1998). The FASB is a supporter of the expensing method when accounting for research

and development costs. And because there was an announcement by the FASB and IASB that

they are working on a convergence project between the US GAAP and IFRS, it is useful for

the debate to determine what the determinants are for US firms to capitalize or expense

research and development costs. Illustrating that software development costs capitalization is

motivated by incentives to manipulate earnings would support the position of the FASB

which supports the expensing alternative. However, illustrating that firms in the US do not

use capitalization of software development costs would support the position of the IFRS

which tend to capitalize more expenditures than US GAAP does. However this study does not

include direct proxies for earnings management, the results found could have potential

influences on the debate concerning the convergence of US GAAP and IFRS in the US.

4. Hypothesis development and research design

This study analyses the determinants of the choice to capitalize or to expense software

development costs. It provides information about the reasons that firms have when selecting

their method of accounting for software development costs. This study may offer insights

with respects to how US firms handle the capitalization versus the expensing choice. This

may have implications on the debate that is ongoing about the harmonization of accounting

standards. Presently, the FASB and International Accounting Standards Board (IASB)

announced that they are committing to the convergence of US and international accounting

standards. Under IFRS, firms have more options to capitalize certain expenses than under US

GAAP; one of the only instances where firms reporting under US GAAP can capitalize their

expenses, is when firms use SFAS No. 86.

In this study, it is assumed that the advantage of capitalizing above expensing software

development costs, is to be able to communicate certain information about the R&D assets of

firms. Furthermore, it was found by McGee (1984) that management could prefer

capitalization above expensing, because bank officers and analysts in the US responded better

to higher profits obtained by capitalizing. Capitalizing implies that a R&D expense is not

immediately subtracted from the profit, but is subtracted in portions, realizing a higher net

profit. However, a study from the UK by Goodacre & McGrath (1997) found that analysts

were not misled by the higher earnings reported for the company capitalizing instead of

14

expensing R&D expenditures. This may be contributed to the differences between the UK and

the US in investor and analyst climate. In the US, firms tend to expense instead of capitalize

their expenses, because there is only one exception to the rule, and that is SFAS No. 86. This

is further demonstrated by a study done by Graham et al. (2005) who finds in his survey that

financial executives find it to be very important that they meet earnings targets, and analysts’

forecasts. Moreover, these financial executives responded that they would be willing to

conduct real earnings management to achieve this. An example of real earnings management

would be postponing R&D expenditures to achieve a positive profit in the current financial

year. If the risk of capitalizing is too high, then this could be a factor that would dissuade US

firms from capitalization. Another factor that could do this, are the costs of measuring the

constructs contained in SFAS No. 86, which Oswald (2002) notes are costs that could keep

firms from capitalizing their R&D expenses.

As empirical proxies for determinants for the choice to capitalize versus expense

software development costs, this study uses six variables in line with Oswald (2002). These

are earnings volatility, profitability, firm size, leverage, R&D intensity, and market-to-book

value. For the first hypothesis, it is expected that earnings volatility will have a positive effect

on the capitalization choice, because it is expected to be an effective way for them to signal

the success of their software development programs; capitalization may reduce the earnings

volatility of the capitalizers. Markarian et al. (2008) also offers insight on this. This study

states that capitalizing research and development costs can be used to smooth earnings, which

implies that managers of firms would try to use the capitalizing option to create a consistent

flow of earnings year after year.

According to theory, managers of firms could have two possible aims for the use of

capitalization of research and development costs in combination with earnings smoothing.

Firstly, there is a positive view of earning smoothing, which states that managers try to

transmit private information to investors, because it is too costly to transmit this information

through other means to the market. This has been researched by Trueman & Titman (1988)

who state that there are many studies that try to focus on the existence of income smoothing

by managers of firms, but that there has been little exploration of why managers of firms may

want to smooth their firm’s earnings.

They find a rational reason of why managers might try to do this by researching firms

in a market setting as it allows for an analysis of the effects of earnings smoothing on stock

prices. They find that managers use income smoothing to reduce the estimates of claimants of

firms about the volatility of its underlying earnings process. This lowers their assessment of

15

the chance of default, thereby being positive for the firm itself and its shareholders. This

lowering of the assessment of the chance of default of the claimants of the firm leads to lower

costs of borrowing and it favorably affects the relations between the firms and its customers,

employees, and suppliers. However, this result is hampered by two assumptions that the

authors make in this study. Firstly, they assume that a firm has the discretion to move income

from one period to the other; there are no blockades which depend on the nature of the

operation of the firm. Secondly, they make an assumption that claimants of the firm cannot

fully observe every operation of the firm and therefore are not fully informed about its ability

to use flexibility to move earnings from one period to another. And finally, it is assumed that

there are no costs pertaining to the smoothing of a firm’s income. The authors state that these

costs could consist if an early recognition of income for the purpose of financial reporting

could also end in an early recognition of a firm’s tax liability. This could potentially result in

a higher value of taxes to be paid, because if tax is recognized prematurely, then the present

value of a tax liability could be higher, pertaining to the effect interest has over time.

Furthermore, Tucker and Zarowin research whether income smoothing improves

earnings informativeness. Just like the previous study conducted by Trueman & Titman, they

do not try to research whether earnings smoothing actually exists, but they assume that

income smoothing is present in a part of the sample, and try to compare the part of the sample

where income smoothing is present and the part of the sample where income smoothing is not

present. They then examine whether income smoothing deteriorates the information can one

could get from earnings, or improves the informativeness of past and current earnings about

future earnings and cash flows. Additionally, they use a new approach where they measure

income smoothing by the negative correlation of a firm’s change in discretionary accruals

with its change in pre-managed earnings. This results in their findings where they find that the

change in present stock price of firms that use a higher proportion of earnings smoothing than

other firms contains more information about the future income than does the change in the

stock price of firms that use a lower proportion of earning smoothing than firms that use a

higher proportion of earnings smoothing. They additionally show that their results are robust

to various variables such as firm size, future growth opportunities, and future income

volatility. These studies show that managers can use earning smoothing as an informative

device as is assumed in our first hypothesis where we expect that earnings volatility will have

a positive effect on the capitalization choice, because they use this capitalization device to

convey information to the market to signal that their software programs are successful; and by

reducing the volatility of earnings it means that investors will have a more positive view of

16

the firm because the risk is perceived to be lower by investors. This assumes that earnings

volatility is interpreted as an important measure of the overall risk of a firm and has an

adverse effect on the value of a firm’s shares.

However, there is also a second possible aim that managers of firms could use when

smoothing earnings by capitalizing software development costs. This is called the negative

view of earning smoothing which states that managers of firms use the smoothing of earnings

to manage earnings. One of the common definitions of earnings management which is used in

the literature on earnings management is stated by Healy & Wahlen (1991):

“Earnings management occurs when managers use judgment in financial reporting and in

structuring transactions to alter financial reports to either mislead some stakeholders about the

underlying economic performance of the company or to influence contractual outcomes that

depend on reported accounting numbers”

What differentiates this negative view of earnings smoothing from the positive view, is that

stakeholders are misled here by the managers of the firms about the underlying economic

performances of the firm or to influence contractual outcomes that depend on reported

accounting numbers to obtain some private gain for the manager himself, while it would

potentially negatively impact the stakeholders. One could give an argument that under the

positive view of earnings smoothing stakeholders are being misled too, and managers would

profit from the earnings smoothing too by, for example, getting a higher bonus as

demonstrated by Healy (1985) who finds that managers make income-increasing accounting

decisions to receive higher bonuses. However, the intention that managers of these firms have

is to communicate the true underlying economic performances of the firm, which is useful for

investors in making capital-allocating decisions. Whereas under the negative view of earnings

smoothing, managers are misleading investors and could potentially let investors make bad

capital-allocating choices, while profiting themselves from it.

If we summarize the information above, and the assumption is made that firms in

industries that are subject to high research and development intensity, which is a reasonable

assumption to be made when researching a high-tech industry. Then we can assume that firms

in this sector are subjected to a high earnings volatility due to the expenditures for R&D,

therefore firms need to communicate the success of their R&D programs to the market, and

firms can do that by lowering the volatility of their earnings, and thus they need to capitalize

earnings to lower earnings volatility. Then we can make the following hypothesis:

17

H1: Ceteris paribus, earnings volatility is positively related to the choice to capitalize

software development costs.

Moreover, for the second hypothesis, it is expected that firms with a lower profitability would

capitalize their software development costs, because it reduces their profit less than expensing

would. This hypothesis is once again based on the positive view of earnings smoothing. A

piece of theory that directly addresses this hypothesis is the zero earnings theory. Where it is

assumed that people concerned with a firm’s viability and profitability, such as customers,

suppliers, lenders, and investors have a certain mentality which makes them have certain

thresholds for earnings. This theory is described by Keane & Runkle (1998) where the authors

state that individuals have a tendency to view continuous data in a discrete form, thereby

having a tendency to divide the world into categories. So in this instance it would mean that

investors view a profit as either low or high, or negative or positive. If we used positive

earnings smoothing as an explanation of why firms would capitalize software development

costs, then it would be a similar reason as used when substantiating the first hypothesis. The

managers of the firms in the sample would like to signal the success of their software

development programs to the market. As investors see a negative profit or a lower profitable

as something that is negative, it would mean that the capital would not flow to the “good”

firms if these firms were not to signal their success by using the capitalization of software

development costs. Therefore, managers of firms have an incentive to signal to investors

what these good firms are. By using capitalization of software development costs, managers

can use the flexibility to manage earnings to improve the allocation of capital in the economy.

On the other side, the more profitable firms, or firms with a positive profit are already

signaling to the market that their software development projects are successful, by having the

merit that their profit is higher, or positive; so they do not need to capitalize their software

development costs. However, when viewed from the negative view of earnings smoothing,

one would expect that firms with a lower profitable, or negative earnings would manage

earnings upward to misinform investors and get positive profits, while managers would

manage the earnings of the firm downward if the firm is profitable, thereby having a higher

chance of a positive profit next year, because earnings are managed by using accruals which

turn around at a certain point (Healy, 1985).

If we summarize the information above, and if we assume that managers of firms use

the capitalization of software development costs to signal to the market that their software

18

development projects are successful, and thus is used to increase earnings informativeness for

investors; then the following hypothesis can be made:

H2: Ceteris paribus, earnings sign is negatively related to the choice to capitalize software

development costs.

Furthermore, for the third hypothesis, it is expected that larger firms would be more willing to

capitalize their software development costs, because the recordkeeping costs are expected to

be more cumbersome for smaller firms than for larger firms. Therefore the third hypothesis

can be noted as:

H3: Ceteris paribus, firm size is positively related to the choice to capitalize software

development costs.

Additionally, R&D intensity is added to research whether the magnitude of R&D

expenditures has an effect on the choice to capitalize software development costs versus

expense them. A reason why this relationship might exist, is because managers of companies

who are R&D intensive will capitalize their successful R&D projects to show the market that

their programs were succesful. This argument is reasonable, because firms that are engaged in

a high proportion of R&D activities would find it costly to directly communicate the contents

of their projects to the market, because it is propriortary information. Disclosing via

capitalization provides an indicator of success without communicating propriatary

information to the market. However, then it is assumed that firms who are not very R&D

intensive, will prefer to communicate their information on R&D projects through other

mediums, because the costs too capitalize are too high in relationship to other methods.

However, Oswald (2002) notes that there are also R&D intensive firms that tend to have

experience with certain statistics that summarize information of the success of their R&D

projects. In that way, it could be less cumbersome for the firm to disclose information through

these summary statistics than to capitalize their R&D projects. However, this study is

focussed on the US computer programming and prepackaged software industry, therefore it is

expected that the industry is very R&D intensive. Therefore, the sign of the relationship

between the choice to capitalize or to expense software development costs and the R&D

intensity of the firms is hypothesized, assuming that the argument on summary statistics is not

19

substantial. If one would summarize the above into a hypothesis, one would get the following

hypothesis:

H4: Ceteris paribus, R&D intensity is positively related to the choice to capitalize software

development costs.

The leverage variable is included, because it is found that managers of more leveraged firms

have more incentives to use capitalization as it offers a way to make income-increasing

accounting decisions. Debt covenants are intended to restrict the managers of firms from

making investment choices and financing decision choices that would result in a reduced

value of the claims that debtholders have over the firm. To the extent that accounting rules

give managers of firms flexibility to make income-increasing accounting decisions, managers

have the opportunity to choose certain procedures written up in the accounting standards that

can allow them to avoid violating debt covenants that they can be engaged in with a lender.

Managers want to avoid debt covenants as they can be costly for the firm, as documented by

Sweeney (1994) who found evidence of lenders requiring concessions from the borrowers to

resolve the default of the loan. She finds evidence that managers of firms who are

approaching debt covenants respond with accounting changes that are earnings-increasing.

She does this by examining accounting changes, costs of default, and covenants that are based

on accounting numbers which are violated by 130 firms that reported these violations in their

annual reports. This argument is substantiated by a study done by Defond & Jiambalvo (1994)

who examine abnormal total and working capital accruals of firms that reported debt covenant

violations in their annual reports. They hypothesize that these covenant restrictions influence

the accounting choice, and therefore they research the accounting years preceding the

covenant violation, and the year after the debt violation. They firstly find that abnormal

accruals are used in the year preceding the covenant violation, and thus this finding can be

used to substantiate the debt covenant hypothesis. They report that firms make

income-increasing accounting decisions when they are close to violating a covenant from a debt

contract, but they also report that even when violation of the debt covenant is certain, they

still make income-increasing decisions to solidify their position in the process when the

debtholder will require concessions from the borrowers. However, it is found that the

evidence in favor of the debt covenant hypothesis is mixed (Markarian et al., 2008). One of

the reasons could be that the studies that examine this hypothesis make use of firms in their

sample that have been found to have violated the debt covenant in their debt contract.

20

However, there is no control group in their samples, making the samples have a selection bias

(Defond & Jiambalvo, 1994), (Sweeney, 1994). It is also stated that firms do not actively

disclose their debt covenants, and that research into these covenants is therefore quite costly.

A variable used as a proxy to do research on the debt covenant hypothesis is the

leverage ratio. There are several studies that do research on the viability of using the leverage

ratio as proxy for debt covenant existance or tightness. Duke & Hunt III (1990) examine a

number of proxies used as proxy for debt covenant existance or tightness. They test the

validity of these proxies by examining its relation to actual debt covenant restriction from a

random sample of US firms. Compared to other studies, this study actaully has a sample of

random firms to solve the problem of the selection bias which are present in other studies

researching the debt covenant hypothesis. The findings of the study supports the use of the

leverage ratio as a proxy.

So based on the information above, if we were to translate this into a hypothesis, one

would expect that managers of more leveraged companies would have more incentives to

capitalize software development costs to make earnings-increasing changes in total income.

This capitalization would loosen restrictions of debt covenants because of two possible

reasons. Firstly, if it would be the case that the debt covenant would be based on profitability

numbers, then capitalization would increase income, because the software development costs

that would be capitalized, would be split over multiple years instead of expensed directly.

Secondly, if the debt covenant would be based on the leverage ratio, then capitalization would

lower this ratio, as capitalization would increase the income, and thus the equity of the firm.

However, there is also evidence that this relationship is ambigious as Oswald (2002) notes

that there is evidence that intangible assets are excluded from public debt contracts.

Therefore, there is no prediction in which direction the relationship will be and it is not clear

whether there is a relationship at all, so the variable will be included as a control variable.

Finally, the market-to-book value of a firm is included as control variable for future growth

opportunities. It was found that the market-to-book value is a sufficient indicator of future

return on equity and thus can be used as a control variable for future growth opportunities

(Penman, 1996).

If the information about the variables used above is summarized into a probit equation,

we will get the following:

CAP.

𝑖𝑡

= 𝛽.

0

+ 𝛽.

1

𝐸𝐴𝑅𝑁

_{𝑉𝑂𝐿𝑖𝑡}

+ 𝛽.

2

𝐸𝐴𝑅𝑁

_{𝑆𝐼𝐺𝑁𝑖𝑡}

+ 𝛽.

3

𝑆𝑖𝑧𝑒.

𝑖𝑡

+ 𝛽.

4

𝐿𝐸𝑉.

𝑖𝑡

+

𝛽.

₅

𝑅𝐷𝐼𝑁𝑇.

_𝑖𝑡

+ 𝛽.

₆

𝑀𝐵.

_𝑖𝑡

+ 𝜀.

_𝑖𝑡

21

CAP.

𝑖𝑡

is a dummy variable which is equal to one if firm i is a capitalizer in year t, zero

otherwise. 𝐸𝐴𝑅𝑁

_{𝑉𝑂𝐿𝑖𝑡}

is the earnings variance of firm i in year t, and earnings variance is

measured as the variance of the firm’s earnings per share divided by the share price.

𝐸𝐴𝑅𝑁

_{𝑆𝐼𝐺𝑁𝑖𝑡}

is a dummy which is equal to one if firm i in year t if profit is positive, zero

otherwise. 𝑆𝑖𝑧𝑒.

_𝑖𝑡

is calculated as the log of assets of firm i in year t. 𝐿𝐸𝑉.

_𝑖𝑡

is calculated as

total liabilities divided by total book value of equity for firm i in year t. 𝑅𝐷𝐼𝑁𝑇.

𝑖𝑡

is

calculated as R&D expenditures divided by total assets for firm i in year t measured at fiscal

year-end. 𝑀𝐵.

𝑖𝑡

is calculated as total market value divided by total book value of equity for

firm i in year t. Finally, 𝜀.

_𝑖𝑡

is the residual for firm i in year t. A probit regression is used,

because the dependent variable is binary (0/1). Firms can either capitalize their software

development costs, or they can expense them, or use a combination of expensing and

capitalization.

5. Data and sample description

The sample initially acquired from Compustat North America contained 1336 unique

firms, and 7577 unique firm-years. This sample was acquired from Compustat using

indicators for industry called SIC codes. The firms with SIC codes that are used are those that

use SIC codes 7370-7372, indicating that they are from the computer programming and

prepackaged software industry. According to Mohd (2005), these industries provide a good

picture of the firms which are obligated to follow SFAS No. 86. The period over which the

data is collected, runs from 1998 to 2006. This data is selected, because the computer

programming and prepackaged software industry makes use of the rules contained in SFAS

No. 86. Additionally, data for total assets, total equity, total liabilities, total market value,

R&D expenses, earnings per share, share price, and capitalized software were required. This

study also requires a minimum of two observations per firm to make an estimation of earnings

volatility. However, due to missing data, some observations had to be excluded from the

sample. This gives a final sample with 5448 unique firm-years with 134 capitalizers, and 5314

expensers. The original sample contained 151 capitalizers, and 7426 expensers. This means

that 2129 observations had to be excluded from the sample due to missing data in one of the

categories used, which are divided in 17 excluded observations pertaining to capitalizers, and

2112 observations pertaining to expensers. An observation is called an expenser observation if

in a year a firm does not capitalize software development costs, and it is called a capitalizer

observation if a firm capitalizes software development costs. The first thing what one notices