• Ingen resultater fundet

CHAPTER V - DATA AND METHODOLOGY

5.2 Data selection

Our data basis is annual reports for Danish portfolio companies bought between 2000 and 2014. In the selection of which portfolio companies that should be included in our sample, the baseline is DVCA’s list. It is a publicly available list that includes every buyout made on the Danish market from 1986 and until today. DVCA is an interest group for venture capital and private equity funds and the list is updated yearly. We are aware of the risk to rely on external material and perhaps the list is not comprehensive, but it is our perception that the list is reliable and objective. The information on the list concerning year for acquiring and selling the portfolio companies and whether or not the PE-fund acquire a majority stake. This is confirmed by scrutinizing all the financial statements for the sample and is therefore not an element that weakens the data material. To obtain a representative and valid sample, we have established six criteria that must be met in order for a portfolio company to be a part of our sample.

52

Figure 7: Strategy for sampling – six criteria’s figure

Source: Own contribution

The first criterion is that the PE-fund must acquire more than 50% of a Danish company between 2000 and 2014. This is crucial because it determines whether or not the PE-fund has controlling interest in the portfolio company and can actively influence the development. If a PE-fund acquires less than 50% of the controlling stakes in a company, we cannot assume that the portfolio company’s direction is solely due to the PE-funds involvement. We have accordingly omitted the transactions that were not majority acquisitions.

To be included in our sample, the portfolio company should be acquired by a “real” PE-fund with a structure described in chapter 2. Companies bought by family funds or investment companies are omitted because they do not have the PE-structure. We are interested in investigating the “PE-effect” and whether or not they are able to outperform their peers. It is not possible to pool PE-funds with family funds and investment companies, because they have different structures and incentive schemes. We wish to isolate the PE-effect and strive to measure performance changes from PE-entry to PE-exit, whereas it is accordingly important not to include secondary buyouts in the sample because there is no change in ownership structure. We however include portfolio companies that are being sold to another PE-fund in the exit-stage in our data material, because it is the change in ownership structure when the first acquisition finds place and the following progress we are interested in measuring relative to peer performance.

Majority investment in a danish company between 2000 and 2014

Portfolio company acquired by a ”real” PE-fund

Change in ownership structure (no secondary buy-outs)

Stand-alone investment (no add-on companies)

Annual reports available in our time frame (group reports)

Declaration of revenue

53 Figure 8: Ownership structure considerations

Source: Own contribution

A secondary buyout is allowed in the last stage of our framework as illustrated in the figure above. The portfolio company should have another ownership structure than PE in the entry-stage of our model. If the portfolio company have not been under PE-ownership when entering our sample, but are being sold as a secondary buyout, they are still included in our sample because we get the full comparison possibility under the first PE-ownership period where the change in ownership structure finds place.

We have also excluded add-on investments in our sample as a stand-alone portfolio company, because it is difficult to measure PE-influence on a stand-alone basis when the add-on company is being consolidated into an existing business. We instead measure on the acquiring company if it fits within our other requirements. We do however allow add-on investments by an existing portfolio company to grow inorganic and thus measure on the existing companies development. The value creation and possible increase on different parameters in the company is difficult to isolate when added to an existing company.

The basis of our analysis is annual reports. If the annual reports are not available one year prior to the acquisition and one year after the exit (alternatively if a firm goes bankrupt), we are not able to isolate the PE-effect and has to omit the company/acquisition from our sample.

Our observation period goes from one year prior to the PE-acquisition and until (if possible) the first annual report after the portfolio company is sold. With this approach, we have the latest result prior to the ownership change and entrance of the PE-fund and the “result” when the PE-fund exits their investment and believe they have added the value they could. In events of bankruptcy, we, however, take the newest possible annual report prior to bankruptcy and include in the sample, to reduce bias towards a positive direction. If the companies that could not deliver an annual report t+1 year after PE-exit would be excluded from the sample, all the bankruptcies would be excluded and the sample would be biased. It is important to mention that we only use group annual reports to evaluate the consolidated (parent) result and not

54

holding company data, as holding companies are closed down after exit whereas parent companies are long lasting.

At last, we require that the portfolio companies report revenues because many of our statistical comparison tests are based on margins such as EBITDA-margin that require a revenue number for the calculation. We are aware of the bias it might create towards a sample for larger companies, because it is often smaller companies that are not required to submit revenue information. We will, on behalf of this, have difficulties on commenting on PE-ownership of smaller companies (often accounting class A+B).

After discarding the minority investments in portfolio companies, there are 151 companies left between 2000 and 2014. The sample is of satisfactory size until the last criterion, where we exclude 35 companies because they do not report revenue and end up with a sample on 43 companies. This is not a satisfactory size for our statistical tests (Newbold, Carlson, & Thorne, 2013), but compared to similar tests made within this scientific area our sample size is tolerable. We treat our final sample as an approximation of a random sample even though it has a statistical uncertainty because of few observations.

Figure 9: Breakdown of sample selection

Source: Own contribution with base from data sample