*2.1. Variable Selection*

Three dependent variables in the measurement of profitability were selected: Return on Equity (ROE), Return on Assets (ROA), and Return on Capital Employed (ROCE). ROE implies the average annual return generated for the equity owners, ROA is the return generated concerning the total assets in the firm and an indicator of how efficiently the company is using its assets. ROCE is a measure for comparing companies in capitalintensive industries (with a lot of debt), as it indicates how well a company is using its overall available capital. The definitions used for the three dependent variables are as follows:

ROE = [Net Income + Taxes]/[Average Stockholders' Equity] (1)

$$\text{ROA} = \text{[Net Income} + \text{Taxes]} / \text{[Average Total Assets]} \tag{2}$$

ROCE = [Net Income + Taxes]/([Average Total Assets]-[Current liabilities]) (3)

The independent variables were selected in such a way that they include both firmspecific and industry-specific determinants. The variables were selected based on earlier choices made in the previous literature. The "net income" used as a control variable includes the effect of taxes (net of tax) in order to eliminate company-specific efforts to minimize taxes. The firm variables were retrieved from the Amadeus database and the industry variables from selected databases (see Table 3). The annual average Feed-in-Tariff rates for solar-, biomass-, geothermal-, wind-, and hydro-energy are studied for the effect on profitability for each one of the studied years.


**Table 3.** Selected independent variables, their specifications, and the source of data. Explanations: \* Annual percentage growth rate of GDP per capita in €. GDP per

Two of the independent variables, [TotalAssets] and [Sales], were log-transformed to make them approximately follow the normal distribution required in the statistical analysis. From the industry-specific determinants, the "change from the previous year's share of RE in electricity consumption" was chosen as a proxy for industry growth. Market concentration was also added as a growth rate, "a percentage increase from the previous year's share of the largest electricity generator in the industry". We found it important to study the separate effects of the company's size in terms of assets and in terms of sales, as well as the growth in sales and assets, as these variables have different implications.

As it is not possible to acquire the amount of FIT-support received by individual companies from public databases, an attempt was made to include them in the quantitative analysis model and to test whether they (partially) explain the variance. The average FIT used in the analysis is an aggregate mean of the average annual FIT received by all the RE sectors.

The selected variables that had a significant correlation with some another independent variable were removed and only one variable from such a pair was kept in the analysis. For the purposes of this research, the variables with a significant correlation larger or equal to ±0.6 to another independent variable were removed from the analysis. More specifically, in the SME data, a strong and statistically significant (5% level) positive correlation of 0.89 between the leverage variables D/A and D/E was found, thus D/E was excluded from the analysis of the SMEs. A strong negative and statistically significant correlation (−0.64) between the Electricity price (Elecpriceh) and the growth rate of the Share of Renewables in Electricity consumption (Elecreshare\_G) was found. Electricity price also correlates strongly with the growth rate of Electricity consumption (ElecCons\_G) (−0.62) and the annual average Feed-in Tariff price (Fitavg) (−0.77); thus, the variable Electricity price was removed from the analysis in both data sets.
