3. Data and research design

To explain the sample selection of technology-firms from the NASDAQ market index, we firstly used the COMPUSTAT database to generate a list of firms that were active on the NASDAQ market as of January 2007. The rationale behind selecting the NASDAQ market is that its constituent members are heavily weighted towards the technology sector, a decision similarly applied in reference to [4]. As we identify the pinnacle of the GFC to be the latter half of 2008, we generate a shortlist which comprises of firms which were active as of January 2007 to ensure that each firm within our final sample have been active for at least a year. We also ensure that all firms within our final sample were active as of third quarter of 2008. Type Variables Description

Also known as the earnings yield, the EP ratio is the inverse of the PE ratio measuring the earnings per share of a company divided by the share price of the company. Generally, the higher an EP ratio gets, the company can either be seen by the market with low confidence in its

value of a firm relative to its market value. This ratio is also used in valuation practices to identify undervalued, typically over 1.00, or overvalued securities, if less than

revenues. It is used to indicate the value attached to each dollar of a company's sales or revenues. Lower values tend to suggest undervaluation while higher values may indicate overvalued stocks or those with higher confidence

Market capitalisation is derived by calculating the company's shares outstanding multiplied by the current market price. It is used as a close proxy for size of the firm,

and typically larger firms tend to survive crises

revenue year on year in Q4 of a particular year. Those with higher growth are assumed to enjoy greater return prospects and lower likelihood of failure

An accounting item that acts as a strong indicator of a company's financial performance. It is useful in displaying the earning potential of a business as well as a gauge to analyse profitability that eliminates the effects of financing

of a company's goods or services. Importantly, the technology sector is amongst the highest users of RD and the level may be viewed by market participants as either an expense or investment initiative, a key part to our study

5-year performance measured by stock returns. These explain the firm's performance from the short to long term

These are the dummy variables assigned 1 if the firm fails and 0 if otherwise in 6-month, 1-year, 3-year, and 5-year

Along with non-discretionary accruals, they make up the total accruals practiced by the company. Discretionary accrual is associated with management choices whereas non-discretionary accruals originate from business conditions. DA is a good proxy for earnings quality and our study employs the Jones and Modified Jones Model for its

The Global Competitiveness Report (2008–2010)

earnings, or is currently undervalued Book to market ratio (BM) The book to market ratio is a ratio that calculates the book

Price to sales ratio (PS) Another ratio that divides a company's stock price by its

assigned by the market

Sales growth (Salesg) Sales growth in our study is defined as the growth in sales

and accounting decisions R&D expenditure (R&D) Refers to expenses linked to the research and development

3m, 6m, 1y, 3y, 5y These are the firm's 3-month, 6-month, 1-year, 3-year, and

ETHICS The scores of firm's ethical behaviour in US collected from

calculation

The table shows the description of the variables we use. It shows the description for our valuation (EP, BM, PS), accounting (MktCap, Salesg, EBITDA, R&D, (3m, 6m, 1y, 3y, 5y) returns, (Failed\_6m, Failed\_1y, Failed\_3y,

1.00

The Roles of Accounting Valuations and Earnings Management in the Survivorship of Technology…

Valuation Earnings to price ratio (EP)

DOI: http://dx.doi.org/10.5772/intechopen.85395

Accounting Market capitalisation (MktCap)

> Earnings before interest, taxes, depreciation, and amortisation (EBITDA)

Failed\_6m, Failed\_1y, Failed\_3y, Failed\_5y

Discretionary accruals

(DA)

Failed\_5y) failures, DA, and ETHICS).

Earnings management

Qualitative data

Variable description.

Table 1.

65

After generating the shortlist of firms, we further break down our sample into defined Industry Classification Benchmark (ICB) codes within the NASDAQ index to arrive at a total of 350 firms. These represent the chapter's scope in evaluating technology firms. The earliest starting point of our data is selected as the yearending 2007 so that we may calculate figures which require a year-on-year change, such as the calculation of yearly sales growth. In other words, if the financial data involves a year-on-year change, the earliest data point used to calculate this change figure will be the 2007 value. All financial and valuation data are generated from the financial year end of 2008, which is December. This is the same for 2009 and 2010.

In addition, we collect the overall firms' ethical behaviour scores (ETHICS) in US from The Global Competitiveness Report (2008–2010). This variable measures the effect of ethical behaviour of the technology firms to their returns and bankruptcies in our further analysis. The descriptions of the accounting valuation and earnings management metrics used in the study are detailed in Table 1.

In terms of their data extraction, we use yearly data where it refers to an accounting reported figure (Sales\_g, EBITDA, RD, and calculation of DA) and quarterly data for valuation metrics (EP, BM, PS and MktCap) from the Bloomberg Terminal ranging from 2007 to 2015. Where the variables are extracted on an annual basis, we use the base year figures 2008, 2009, and 2010. Where the variables are quarterly, we use the Q4 data from the base years, 2008, 2009, 2010 to form the independent variable pool as this aligns with the financial year-end figures of our accounting data.

All accounting items required to conduct the DA procedures are extracted using yearly data from the Bloomberg database for each year, 2008–2010. The 2007 figures are also generated considering that some items require a calculation to measure the change. The non-discretionary accruals using the Jones Model is express as:

$$NDA\_t = a\_1 \left(\frac{1}{A\_{t-1}}\right) + a\_2(\Delta REV\_t) + a\_3(PPE\_t) \tag{1}$$

The total accruals using the Jones Model is express as:

$$TA\_t = a\_1 \left(\frac{1}{A\_{t-1}}\right) + a\_2(\Delta REV\_t) + a\_3(PPE\_t) + \theta\_1 \tag{2}$$

where ΔREVt = revenues in year t less revenues in year t � 1 scaled by total assets at t � 1; PPEt = gross property plant and equipment in year t scaled by total assets at t � 1; At�<sup>1</sup> = total assets at t – 1; a1a2a<sup>3</sup> = firm-specific parameters; TAt = total accruals scaled by lagged total assets; income before extraordinary items—CF from continuing operations.


The Roles of Accounting Valuations and Earnings Management in the Survivorship of Technology… DOI: http://dx.doi.org/10.5772/intechopen.85395

The table shows the description of the variables we use. It shows the description for our valuation (EP, BM, PS), accounting (MktCap, Salesg, EBITDA, R&D, (3m, 6m, 1y, 3y, 5y) returns, (Failed\_6m, Failed\_1y, Failed\_3y, Failed\_5y) failures, DA, and ETHICS).

#### Table 1.

Variable description.

3. Data and research design

of our accounting data.

express as:

64

To explain the sample selection of technology-firms from the NASDAQ market index, we firstly used the COMPUSTAT database to generate a list of firms that were active on the NASDAQ market as of January 2007. The rationale behind selecting the NASDAQ market is that its constituent members are heavily weighted towards the technology sector, a decision similarly applied in reference to [4]. As we identify the pinnacle of the GFC to be the latter half of 2008, we generate a shortlist which comprises of firms which were active as of January 2007 to ensure that each firm within our final sample have been active for at least a year. We also ensure that all firms within our final sample were active as of third quarter of 2008. After generating the shortlist of firms, we further break down our sample into defined Industry Classification Benchmark (ICB) codes within the NASDAQ index to arrive at a total of 350 firms. These represent the chapter's scope in evaluating technology firms. The earliest starting point of our data is selected as the yearending 2007 so that we may calculate figures which require a year-on-year change, such as the calculation of yearly sales growth. In other words, if the financial data involves a year-on-year change, the earliest data point used to calculate this change figure will be the 2007 value. All financial and valuation data are generated from the financial year end of 2008, which is December. This is the same for 2009 and 2010. In addition, we collect the overall firms' ethical behaviour scores (ETHICS) in US from The Global Competitiveness Report (2008–2010). This variable measures the effect of ethical behaviour of the technology firms to their returns and bankruptcies in our further analysis. The descriptions of the accounting valuation and

Accounting and Finance - New Perspectives on Banking, Financial Statements and Reporting

earnings management metrics used in the study are detailed in Table 1.

In terms of their data extraction, we use yearly data where it refers to an accounting reported figure (Sales\_g, EBITDA, RD, and calculation of DA) and quarterly data for valuation metrics (EP, BM, PS and MktCap) from the Bloomberg Terminal ranging from 2007 to 2015. Where the variables are extracted on an annual basis, we use the base year figures 2008, 2009, and 2010. Where the variables are quarterly, we use the Q4 data from the base years, 2008, 2009, 2010 to form the independent variable pool as this aligns with the financial year-end figures

All accounting items required to conduct the DA procedures are extracted using

þ a2ð Þþ ΔREVt a3ð Þ PPEt (1)

þ a2ð Þþ ΔREVt a3ð Þþ PPEt ϑ<sup>1</sup> (2)

yearly data from the Bloomberg database for each year, 2008–2010. The 2007 figures are also generated considering that some items require a calculation to measure the change. The non-discretionary accruals using the Jones Model is

where ΔREVt = revenues in year t less revenues in year t � 1 scaled by total assets at t � 1; PPEt = gross property plant and equipment in year t scaled by total assets at t � 1; At�<sup>1</sup> = total assets at t – 1; a1a2a<sup>3</sup> = firm-specific parameters; TAt = total accruals scaled by lagged total assets; income before extraordinary

1 At�<sup>1</sup> 

NDAt ¼ a<sup>1</sup>

TAt ¼ a<sup>1</sup>

items—CF from continuing operations.

The total accruals using the Jones Model is express as:

1 At�<sup>1</sup> 

We also compute the non-discretionary accruals using the Modified Jones Model which is express as:

$$NDA\_t = a\_1 \left(\frac{1}{A\_{t-1}}\right) + a\_2(\Delta REV\_t - \Delta REC\_t) + a\_3(PPE\_t) \tag{3}$$

we conduct logistics regression using 3-models for each base year measuring 1-year failure, 3-year failure, and 5-year failure. One year is chosen as the earliest measure of failure as any earlier, such as 6-months, would consist of an insignificant number of firms that have failed. Logistic regression results earlier than 1-year, in

Our linear regression model for determining future performance is express as:

The Roles of Accounting Valuations and Earnings Management in the Survivorship of Technology…

rat ¼ a þ b1EP þ b2BM þ b3PS þ b4MktCap þ b5EBITDA

And our logistic regression model for determining future failure is express as:

In Eq. (4), the methodology for the linear regression involves a similar framework used in reference to [4], examining the market value of securities in the tech sector during the tech-bubble of 2000. In our case, the dependent variable is selected as the percentage return over the prescribed period (3m, 6m, 1 y, 3 y, and 5 y) from the end of the selected base-years (2008, 2009, and 2010). The independent variables include our valuation, accounting, and earnings management variables of each underlying security in the sample, as described in Table 1. A positive significant coefficient would imply a positive relationship between the

Our linear regression model for determining future performance using our

Furthermore, our logistic regression model for determining future failure using

Then the corresponding null and alternative hypotheses for linear (Eq. (6)) and logistic regressions (Eq. (7)) are coefficients equal zero and non-zero, respectively,

Table 3 is the result of the multivariate analyses using the Jones Model and

<sup>4</sup> In this paper, we mostly discuss the results that show significantly strong results (i.e., p-values ≤0.05). Weakly significant results (i.e., 0.05 < p-values <0.10) are sometimes aligned with the strongly

, and follows the traditional relationship with

þ b7MktCap þ b8EBITDA þ b9ETHICS þ b106m… þ b135y

rat ¼ a þ b1EP þ b2BM þ b3PS þ b4MktCap þ b5EBITDA þ b6Salesg

Failed firmt ¼ a þ b1EP þ b2BM þ b3PS þ b4Salesg þ b5MJDA þ b6RD

Failed firmt ¼ a þ b1EP þ b2BM þ b3PS þ b4Salesg þ b5MJDA þ b6RD þ b7MktCap þ b8EBITDA þ b91y… þ b115y

(4)

(5)

(6)

(7)

þ b6Salesg þ b7MJDA þ b8RD

our case, are limited in their validity.

DOI: http://dx.doi.org/10.5772/intechopen.85395

indicator and future returns.

as before.

67

overall data including ETHICS is express as:

þ b7MJDA þ b8RD þ b9ETHICS

our overall data including ETHICS is express as:

4. Results and discussion of findings

significant results, but we do not actively discuss these.

4.1 Future performance

shows that the R&D is significant<sup>4</sup>

where ΔRECt = net receivables in yeart less net receivables in yeart � 1 scaled by total assets at t – 1; NDAt = non-discretionary accruals scaled by total assets of prior year.

Once we have calculated the discretionary accruals value for each firm in their MJDA and JDA forms, we then measure each firm's value relative to the sample median in percentage terms similar to the methodology adopted by past studies, refer to [18]. Consequently, a positive value refers to a firm who has greater discretionary accruals for that particular year relative to its peer group. The descriptive statistics of our overall data is shown in the following Table 2.

#### 3.1 Research design

The core methodologies used for our research include the linear regression and logistic regression models. In our linear regression section investigating H1, we conduct OLS regressions for each base year measuring 3-month, 6-month, 1-year, 3-year, and 5-year performance as the dependant variable. In addressing H2,


The table shows the descriptive statistics of the data set we used. We show the average, median, minimum (min), maximum (max), standard deviation (Std.), 10th percentile (10th per), 90th percentile (90th per), and the total number of sample (N) for our overall data. Our overall data include 3m, 6m, 1y, 3y, 5y, EP, BM, PS, MktCap, Salesg, EBITDA, Failed\_6m, Failed\_1y, Failed\_3y, Failed\_5y, MJDA, JDA, R&D, and ETHICS from 2008 to 2010.

#### Table 2. Descriptive statistics.

The Roles of Accounting Valuations and Earnings Management in the Survivorship of Technology… DOI: http://dx.doi.org/10.5772/intechopen.85395

we conduct logistics regression using 3-models for each base year measuring 1-year failure, 3-year failure, and 5-year failure. One year is chosen as the earliest measure of failure as any earlier, such as 6-months, would consist of an insignificant number of firms that have failed. Logistic regression results earlier than 1-year, in our case, are limited in their validity.

Our linear regression model for determining future performance is express as:

$$\begin{aligned} r\_{a\_t} &= a + b\_1 \text{EP} + b\_2 \text{BM} + b\_3 \text{PS} + b\_4 \text{MktCap} + b\_5 \text{EBITDA} \\ &+ b\_6 \text{Sales}\_{\text{g}} + b\_7 \text{MfDA} + b\_8 \text{RD} \end{aligned} \tag{4}$$

And our logistic regression model for determining future failure is express as:

$$\begin{aligned} \text{Failed } \text{firmt} &= a + b\_1 \text{EP} + b\_2 \text{BM} + b\_3 \text{PS} + b\_4 \text{Sales}\_{\text{g}} + b\_5 \text{MJD} + b\_6 \text{RD} \\ &+ b\_7 \text{MktCap} + b\_8 \text{EBITDA} + b\_9 \text{My} \dots + b\_{11} \text{Sy} \end{aligned} \tag{5}$$

In Eq. (4), the methodology for the linear regression involves a similar framework used in reference to [4], examining the market value of securities in the tech sector during the tech-bubble of 2000. In our case, the dependent variable is selected as the percentage return over the prescribed period (3m, 6m, 1 y, 3 y, and 5 y) from the end of the selected base-years (2008, 2009, and 2010). The independent variables include our valuation, accounting, and earnings management variables of each underlying security in the sample, as described in Table 1. A positive significant coefficient would imply a positive relationship between the indicator and future returns.

Our linear regression model for determining future performance using our overall data including ETHICS is express as:

$$\begin{aligned} r\_{a\_t} &= a + b\_1 EP + b\_2 BM + b\_3 PS + b\_4 M \text{kftCap} + b\_5 EBITDA + b\_6 \text{Sales}\_{\text{g}} \\ &+ b\_7 MjDA + b\_8 RD + b\_9 ETHICs \end{aligned} \tag{6}$$

Furthermore, our logistic regression model for determining future failure using our overall data including ETHICS is express as:

$$\begin{aligned} \text{Failed firm} &= a + b\_1 \text{EP} + b\_2 \text{BM} + b\_3 \text{PS} + b\_4 \text{Sales}\_{\text{g}} + b\_5 \text{MJD} + b\_6 \text{RD} \\ &+ b\_7 \text{MktCap} + b\_8 \text{EBITDA} + b\_9 \text{ETHICs} + b\_{10} \text{бm...} + b\_{13} \text{Sy} \end{aligned} \tag{7}$$

Then the corresponding null and alternative hypotheses for linear (Eq. (6)) and logistic regressions (Eq. (7)) are coefficients equal zero and non-zero, respectively, as before.
