Gender Diversity on the Board - BI Norwegian Business School
Gender Diversity on the Board - BI Norwegian Business School
Gender Diversity on the Board - BI Norwegian Business School
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
GRA 19001 Master Thesis<br />
5 Statistical tests<br />
5.1 Base-case model<br />
Our data set c<strong>on</strong>tains repeated observati<strong>on</strong>s of <strong>the</strong> same firm for up to eight years. We<br />
use <strong>the</strong> fixed effects model to handle this panel data setting. Fixed effects allow us to<br />
c<strong>on</strong>trol for unobserved explanatory variables and to analyze change over time. The more<br />
time points a firm is observed, <strong>the</strong> more informati<strong>on</strong> <strong>the</strong> firm provides. Hence, we <strong>on</strong>ly<br />
study firms that are present for at least three years to ensure that sufficient informati<strong>on</strong> is<br />
provided. The first step in <strong>the</strong> fixed effect model is to compute <strong>the</strong> mean for each<br />
company, for each variable, and subtract this mean from all observati<strong>on</strong>s. For variables<br />
that stay c<strong>on</strong>stant over time, <strong>the</strong> value in each year equals <strong>the</strong> average value across years.<br />
Hence, it cancels in <strong>the</strong> regressi<strong>on</strong> equati<strong>on</strong>. When means are subtracted, we estimate <strong>the</strong><br />
model using ordinary least-squares (OLS). (Petersen, 2004)<br />
Table 5.1 shows results from OLS when estimating <strong>the</strong> base-case model by <strong>the</strong> fixed<br />
effects approach as described above. To c<strong>on</strong>trol for o<strong>the</strong>r firm characteristics that can<br />
influence <strong>the</strong> dependent variable we c<strong>on</strong>trol for firm size and firm age in <strong>the</strong> regressi<strong>on</strong>.<br />
The base-case model will be presented for large n<strong>on</strong>listed firms as well as all n<strong>on</strong>listed<br />
firms. We limit <strong>the</strong> attenti<strong>on</strong> to individual coefficients with a p-value of 10% or less.<br />
22