Research Model

coefficient indicates that an increase in the variable leads to an increase (decrease) in the probability of attaining profitability and a decrease (increase) in the time-to-profitability. A detailed discussion of CPH models is available in Cox (1972). In addition to identifying coefficient signs of variabes, it is also useful to assess the economic impact of these variables by evaluating their impact on the risk or hazard that a currently unprofitable biotech firm will be profitable in the future. For continuous independent variables the hazard ratio represents the estimated percent change in the hazard of the event (attainment of profitability) for a one unit increase in the covariate of interest (controlling for other covariates) and is obtained by subtracting one from the hazard ratio and multiplying by 100. For indicator variables, the hazard ratio is interpreted as the estimated hazard of the event of interest occurring for those with a value of 1 relative to the estimated hazard for those with a value of 0 after controlling for other covariates.

In order to test the linear relationships between variables, this study uses Pearson Correlation and Spearman Correlation to make sure whether there are collinear problems between independent variables.¹

3.4 Research Model

3.4.1 Dependent Variable and Independent Variables

I use quarterly operating income before depreciation as my measure of operating profitability. I define the event in my analysis as the attainment of a quarter of operating profitability for a period of five years after the IPO. In the post-IPO period, firms will either attain profitability, fail, or remain unprofitable until the end of our tracking period.

Censored observations represent IPO firms that are unable to attain profitability by the end of our tracking period. If a firm has a quarter of operating profitability after the IPO then we assign it the profitable status and compute the time-to-profitability as the number of quarters elapsed between the IPO quarter and the quarter for which the firm first reported operating profitability. The construction of the dependent variable is on the basis of combining the time to occurrence of event (profitability) with the dichotomous status variable (attained profitability status versus remains unprofitable at

1 See Allison (2000) for further details on the interpretation of hazard ratios for quantitative and indicator variables. Hellman and Puri (2002) interpret the hazard ratio in a similar manner to that described above for indicator variables.

‧

end of tracking period). The dependent variable in the hazard model, therefore, denotes the likelihood that an Internet IPO firm will attain profitability in each period.

About independent variables, on the basis of previous research (e.g. Welboume and Andrews1996; Cyret al. 2000; Certo et al. 2001), I use twenty different types of risk factors commonly listed in the prospectus as my independent variables. These twenty different types of risk factors are dummy variables in my model. They are coded one if firms disclose them in the prospectors, otherwise, coded zero. Besides, I also use four subgroups of them to represent another kind of classification. They are dummy variables, too. They are coded one if firms disclose them in the prospectors, otherwise, coded zero.

In addition to types of disclosures, I also score them. Score of each item and four subgroups are independent variables, too. I score them in four levels. Score 4 gives to disclosures that provides quantitative information, or the risk factor is clearly defined in monetary terms or actual physical quantities. Score3 gives to disclosures that provides qualitative information specific as toactions, persons, events, or places. Or the impact on the company is clearly evident. Score 2 gives to disclosures mentioned only generally, not specific. Score1 gives to disclosures are immaterial to the financial condition and results of the corporation. In a word, each item has a score, ranging from four to one, the best to the worse.

3.4.2 Control Variables

First, the extant literature has widely recognized the potential for third party certification as a solution to the information asymmetry problem in the IPO market (Beatty, 1989; Carter and Manaster, 1990; Megginson andWeiss, 1991; Jain and Kini, 1995, 1999b; Zimmerman and Zeitz, 2002). The theoretical basis for third party certification is drawn from the signaling models which argue that intermediaries such as investment bankers, venture capitalists, and auditors have the ability to mitigate the problem of information asymmetry by virtue of their reputation capital (Booth and Smith, 1986; Megginson and Weiss, 1991; Jain and Kini, 1995, Carter et al., 1998). In addition to certification at the IPO, intermediaries, through their continued involvement, monitoring, and advising role have the ability to enhance performance after the IPO. I use dummy variables to measure them. If a firm is venture-backed, coded 1, otherwises, coded 0.

Second, I include the number of employees at the IPO (NUMEMP) as a control

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

34

variable. It proxies for the extent of human capital deployed in the IPO firm. Firms operating in the Internet industry have balance sheets that look considerably different from firms operating in more traditional industries because to a large extent they are less dependent on tangible assets and more reliant on intangible assets such as ideas, knowledge, and creativity. Therefore, one of the main assets for technological firms is their human capital base consisting of developers, programmers, designers, and similar knowledge based workers. As such, I expect the likelihood of attaining post-IPO profitability to be positively related to the number of employees.

Third, I include the variable FIRMAGE measured as one plus the age of the firm at IPO as a control variable. Firms that go public prematurely are unlikely to be adequately prepared and financed to withstand the various challenges in the product and financial markets facing newly public firms. Therefore, these firms are less likely to be on the path to achieving profitability compared to firms that had developed sufficiently prior to going public. As such, I expect that the probability of attaining post-IPO profitability is likely to be positively related to the age of the firm at the time of going public.

Finally, consistent with several studies in the IPO literature I include firm size (LSIZE) as measured by the natural logarithm of gross proceeds at the IPO as control variables.

Combine all the factors I use in this research, and I draw the diagram to present clearly.

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

35