The section explains the components embedded in the process of developing the bases for this research. Figure I shows the components of the research onion which includes philosophy, approach, strategies, time horizon and data collection. Research philosophy holds assumptions that indicate overall bases of the research. Sunder et.al (2016) argued, research philosophy as the process of beliefs and assumptions about the development of knowledge. In this paper, assumptions are found closer to positivism, as the research goal is to address possible impacts of the quantifiable variables which are obtained by using factual data. Consequently, the research approach in this context is deductive in nature, because the attempt is made to observe the established concept of risk and return.
Primarily, the time frame of the study is cross-sectional which comprised of ten years, from 2007 till 2016. Observed time horizon includes the era of global financial crises of 2007-8, followed by recovery in the market in later years. However, this research further used mix approach by making five different subsets in order to study the variation in the outcomes of different time spans. Figure II explains the time frame and sampling approach used in the study. Secondary data is used to evaluate the stated relevant hypothesis in this study. Data is collected from the financial statements of 10 US-based passenger airlines listed in NASDAQ. Spirit airline holds business segment portfolio with other US airlines, as the result, partner firms which include delta air, American airlines and southwest paid fuel expense on behalf of Spirit airlines, hence it is excluded from the sample. Finally, 9 airline firms are considered relevant for the study.
As discussed in the previous session, data consists of 9 airline firms for the period of 10 years as it holds the properties of panel data. Hence, it is appropriate to use the advanced econometric technique to determine slope coefficients of observed variables. Regression analysis considered as an important tool to develop the relationship between dependent and independent variables. Traditionally, simple least square method has used in similar studies which would leads to biasness towards the result, especially in panel data study. Because the data used in this study may include unobserved heterogeneity which are firm specific characteristics, therefore simple OLS method is ineffective to include the impact of unexplained heterogeneity existed in the model. As mentioned before, panel data is used in this study which allows to take advantage of GLS by using fixed effect and random effect. In this research, both random and fixed effect methods are used to obtain the slopes of the independent variables. Yang and land (2008) explained Fixed effect eliminates the impact of unobserved heterogeneity which enables the model to reduce the impact of any correlation between individual characteristics and independent variables. Secondly, it keeps slope coefficient persistent for each cross-section within the firms. On the other hand, random effect estimates are considered more effective as it considers time-dependent variable into account. Moreover, standard error in random effect estimation is significantly lower than the fixed effect model. Finally, Hausman test is applied to assess the appropriateness of the estimates obtained by both models.
Assumptions of fixed and random effect models are used to establish the relationship among the variables under the following equations. Equation (i) is drawn by considering the bases of fixed effect model, and equation (ii) is based on the postulates of random effect model.
PROFit = ?0it + ?1FUELit + ?2OEexFUELit + ?3SIZEit + ?4GROWit + ?5RISKit + µit …. (i)
PROFit = ?0it + ?1FUELit + ?2OEexFUELit + ?3SIZEit + ?4GROWit + ?5RISKit + ?it + µit …. (ii)
Variables and proxies
The section focused on the calculation of the relevant variables and proxies used in this study. As previously mentioned, five independent variables are used to estimated potential impact on profitability. Computation of both dependent and independent variables are stated as follows,
Dependent variable PROFit, represents two profitability ratios which are used in this research. ROA (return on assets) and ROS (return on sales) ratios are used separately in the equation (i) and (ii) to evaluate appropriate profit proxy suitable for the equation. Calculation of both ratios are stated below,
ROA = net profit/total assets
ROS = (net profit/sales) – 1
FUELit is considered as an independent variable in the equation which is calculated by dividing total fuel expense from total sales of the firm (i) in a year (t).
Independent variable OEexFUELit is calculated by taking operating expense excluding fuel cost, then divided by total sales of the firm (i) in a year (t).
Proxy of SIZEit of the firm is calculated by taking natural log of total assets of the firm (i) in a year (t).
GROWit represents the proxy of growth, calculated by dividing sales value of current year with base year, followed by the natural log of the residual.
ln (Total Sales1 / Total Sales0)
RISKit is calculated by observing the variation in earnings before interest and tax (EBIT) of firm i in time t. Risk proxy is calculated by taking deviation from current year EBIT value with last year EBIT, then the residual is divided by last year EBIT.
?1 to ?10 are the slope coefficients of the independent variables.
?0, ?0it are the common intercept (random effect) and intercept of firm i in time t (fixed effect) respectively.
?i is the unobserved heterogeneity
µit is the stochastic error term of firm i in time t