INOM
EXAMENSARBETE
TEKNIK,
GRUNDNIVÅ, 15 HP
,
STOCKHOLM SVERIGE 2020
Assessing the Operational Value
Creation by the Private Equity
Industry in the Nordics
ERIK HARRYSSON
ADAM WUILMART
KTH
Assessing the Operational Value
Creation by the Private Equity
Industry in the Nordics
Erik Harrysson
Adam Wuilmart
ROYAL
Degree Projects in Applied Mathematics and Industrial Economics (15 hp) Degree Programme in Industrial Engineering and Management (300 hp) KTH Royal Institute of Technology year 2020
TRITA-SCI-GRU 2020:119 MAT-K 2020:020
Royal Institute of Technology
School of Engineering Sciences
KTH SCI
BSc Thesis: Assessing the Operational Value Creation by the
Private Equity Industry in the Nordics
Erik Harrysson
1and Adam Wuilmart
21Industrial Engineering and Management, KTH. Email:erikharr@kth.se 2Industrial Engineering and Management, KTH. Email:wuilmart@kth.se
Abstract
More and more capital is being directed towards the private equity industry. As a result, private equity owned firms make up an increasingly large share of the economy. Therefore, it is becoming more important to understand the nature of how the operational performance of firms change under private equity ownership. This study looked at how the operational efficiency in terms of EBIT margin changed over a three year period after a private equity acquisition in the Nordic market. The study found that companies which had an initial positive EBIT margin behaved differently from companies with an initial negative EBIT margin and therefore two separate models where created. It was found that in the case where the company had a positive EBIT margin before being bought by a private equity firm saw an average decrease in EBIT margin of 1.14% units. In the case of a firm with initial negative EBIT margin a private equity acquisition led to an average increase in EBIT margin by 1.99% units compared to the reference data. This study thus shows that private equity ownership affects the operational efficiency of companies. Moreover, it shows that one should make a distinction between PE ownership in profitable growth cases and turn-around cases of inefficient companies and that the impact of PE ownership in terms of effect on operational profitability can be vastly different depending on the nature of the acquisition in this regard.
Keywords: Private Equity, Operational Value Creation, Linear Regression, Nordic Market, EBIT Margin, Operational Efficiency.
1
Introduction
1.1 Introducing the Subject Area
As the stock market sees record high multiple valuation, more investors seek alternative ways to allocate their capital to ensure high future yields. Private equity is an investment class, consisting of companies that are not listed on a public exchange. The private equity industry is growing and more companies than ever before are being acquired by private equity firms (Rooney2020).
Private equity firms raise capital from investors into private equity funds. They then use this capital base in combination with high leverage to buy out public or pri-vate companies to become the sole owners. They hold the company during the lifetime of the fund which is usually around ten years. There are different strategies employed by different firms. However, most firms use a combination of acquisition strategies, financial and gov-ernance engineering as well as operational engineering to increase the value of its portfolio companies. During the lifetime of the fund, they take out both performance and management fees from the investors (Kaplan and Strömberg2009).
Data suggests that the Nordic private equity market is especially well developed compared to other coun-tries, where e.g. in Sweden private equity owned
com-panies make up for 6.8% of GDP (Ehrs2020). Since the Nordic private equity market is highly developed and constitutes a large share of the economy, the value cre-ation of private equity-firms is a key factor to ensure continued productivity growth in the region. Therefore, it is of interest to investigate if private equity ownership is good or harmful for the value creation in companies.
1.2 Purpose and Research Question
The purpose of this thesis is to investigate the opera-tional efficiency increase of private equity (PE) owned companies in the Nordics during the time after a PE acquisition. This investigation is undertaken using mul-tiple linear regression analysis on a dataset collected from Orbis. Hence, the research question to be investi-gated is as follows.
"Does private equity ownership increase the opera-tional efficiency of Nordic companies during the years after an acquisition when compared to non-acquired companies?"
the industry and policy makers who are interested in the impact of PE ownership on the quality of Nordic businesses. This question is particularly relevant since PE companies have scrutiny in the media for having a short term perspective on value creation and producing their fund returns through financial engineering (Froud and Williams2007).
To specify the scope, it is necessary to define what is referred to as the "Nordics". The Nordics in this set-ting includes Sweden, Finland and Norway. The focus is primarily on the transactions of Nordic based private equity firms located within these countries. However, transactions within the Nordics by some international firms with extensive operations in the Nordics are also included.
1.3 Previous research
The subject area of whether PE ownership increase or decrease the operational efficiency of companies has been studied quite extensively. One of the most elaborate summaries of the current research situation within the field was presented by Kaplan and Strömberg (2009). The emerging consensus within the field is that leveraged buyouts and particularly management buy-outs both enhance performance and have a prominent positive impact on management practices. However, this does not necessarily mean that the risk-adjusted re-turn to investors from PE funds is higher than that of the stock market net of fees which is under more debate.
One study of particular interest is Viral V. Acharya and Kehoe (2013) which performed a regression analy-sis, similar in methodology to the method used in this article, to measure the effect on operational efficiency from PE ownership in the British market over a decade ago. They found that a private equity acquisition in-crease EBITDA/Sales with an average of 4%.
Lastly, the most relevant study of this kind on the Swedish market is an extensive master thesis from 2007 which looked at post buyout EBITA margin changes to assess the operational value creation of PE funds in the region. They found a positive impact of magnitude 3.07%. (Grubb and Jonsson2007).
This article contributes to the body of research by investigating if the effects still holds more than a decade after the last analysis of the Swedish market and to see if previous findings can be confirmed when analysing a new dataset based on the 2010s.
2
The private equity industry
2.1 Private equity firms
A typical private equity firm is a partnership with a small number of investment professionals who usually have an investment banking background. A standard firm has been found to have approximately 13 employees while the largest firms have around 100 employees.
The private equity firm raise equity capital through private equity funds. Investors commit to provide a cer-tain amount of money as well as to pay a management fee. In general, a fund has a fixed life duration, usually ten years. Common investors are wealthy individuals, public pension funds and endowments (Kaplan and Strömberg2009).
2.2 Acquisition strategy
A central strategic dimension of private equity funds is firm investment focus, both by industry and by stage. It has been found by studying the UK market that fund specialization in terms of industry and stage relative to other PE funds led to higher post buyout profitability. This seems to be a result of an increased ability to men-tor/manage ones portfolio companies and increased ability make suitable acquisition selections when one is focused in terms of industry and stage of development for ones investments (Robert Cressy2007).
Bain has another perspective on acquisition strate-gies in their global private equity report for 2019. They see four different acquisition strategies being executed in the market. The first strategy is "Buy and Build" which means to buy a new company and scale it up by acquir-ing multiple smaller and cheaper competitors result-ing in a form of multiple arbitrage. The bigger firms on the other hand often use a strategy named "Merger-Integration" which in practice means to acquire and ex-ecute large strategic mergers to produce value through synergies and combined operational strength. Lastly, a new strategy gaining momentum is "Advanced An-alytics" where data analytics is used to make better acquisition selections for one’s portfolio. (Bain2019)
2.3 Leveraged buyouts
price. (Kaplan and Strömberg2009).
2.4 Nordic private equity market
The Nordic private equity market is young compared to other European peers. The first Nordic exit by a private equity fund took place in the late 90s (Pedro, Fjellaker, and Camas2012).
Since then the private equity activity in the Nordics has increased significantly. In 2018 the Nordic buy-out funds raised 22.4 billion euro. The region is dom-inated by EQT which raised 10.75 billion euro for the EQT VIII fund. 148 buyouts took place during the year by 72% Nordic firms and 28% non-Nordic firms (Argentum
2018).
The general characteristics of the Nordic private equity market are quite different from the US market. The investment universe is smaller in size, fundraising is more challenging since local firms rely on interna-tional investors and the credit market is dominated by banks with less diversity in financing options than the US credit market. From a cultural perspective manage-ments motivation factors are seen as different than in the US and there is a higher political interest in reducing the tax advantages for private equity firms (Spliid2013).
2.5 Financial and Governance Engineering
There are three main ways in which PE firms increase the value of the companies they have invested in during the holding period. First of all, they generally practice governance and financial engineering which has been common since the late 1980s. This includes improving incentives, which is usually done by increasing the eq-uity stake of the management team fourfold compared to before the LBO. The result is that the management team is rewarded for good performance, but also carry some of the downside risk. Furthermore, their equity stake will be illiquid until the exit which further aligns management interests with those of the owners.
Secondly, a common practice is to increase the lever-age of the company which is usually a direct result of the leveraged buyout (LBO). Increased leverage reduces free cashflow and thus the risk of management wasting money. As a result, the business should be run more effi-ciently. Furthermore, the company will benefit from the tax deductability of interest. However, increase lever-age too much and the benefits will be reduced by the increased risk of financial distress.
Thirdly, PE firms practice governance engineering in terms of how they reduce the size of the board, meet more frequently with managers and take a more active
role in their portfolio companies than the boards of pub-licly companies. One example of the active role PE firms usually take is that PE firms generally do not hesitate on replacing non-performing management (Kaplan and Strömberg2009).
However, the idea that PE firms enhance value cre-ation through governance and financial engineering has received criticism. One main criticism by Leslie and Oyer (2009) is that the increased leverage and higher debt levels disappear shortly after the PE owned com-pany goes public (within 1-2 years). Furthermore, they found it hard to prove that these interventions had any effect on either profitability or operational efficiency.
Some go as far as describing private equity firms as establishing a culture of value extraction in capitalism through over-leverage, instability, asset stripping, a lack of transparency and large rewards for fund managers. Extracting wealth from well-functioning companies in order to benefit a small number of owners and partners (Froud and Williams2007).
However, the studies which made these claims are starting to seem outdated as more and more evidence is built against the view that PE firms strip companies of assets, fire employees and destroy value. One partic-ularly interesting study of 3200 PE target firms between 1980 and 2005 examined the effect of PE ownership on employee count changes. It found that PE ownership led to moderate net job losses, however with large in-creases in gross job creation and destruction. PE firms thus significantly scaled up productive parts of the firm while significantly downsizing parts of the firm with low profitability and productivity (Davis et al.2014).
Another interesting perspective on financial and governance engineering in PE is that private equity firms can gain favourable loans through relationship banking. A study of 1590 loans to private equity firms found that PE firms often achieve favourable loans as a result of the relationships they have built with banks over time (reducing information asymmetry) and the fact that banks price loans to cross-sell other fee busi-ness. One standard deviation increase in bank relation-ship strength and potential for cross selling produces an effect equal to 4% extra return to equity holders in the PE firm because of more favourable loan conditions (Ivashina and Kovner2011).
2.6 Operational Engineering
engineering as one of their main strategies for increas-ing the value of their holdincreas-ings. Today large private equity firms are specialized around industries within which they hire people with operational background in the industry. The operational value creation might take place in the form of repositioning, productivity improvements, strategic changes or cost-cutting, acqui-sition opportunities and management changes (Kaplan and Strömberg2009).
One relevant aspect of the operational engineer-ing of PE firms which has received a lot of exposure is whether the firms influence the operational focus to-wards short term value creation compared to long term value creation. When observing investing in innova-tion as measured by patenting activity it was found that firms involved in private equity transactions showed no significant shifts in the fundamental nature of re-search, the patents were more cited and the research was more focused on the most important areas of re-search (Lerner, Sorensen, and Strömberg2011). This indicates that PE ownership does not lead to less invest-ment in innovation and that it instead might increase the long-term value of the company’s research by mak-ing the process more efficient.
2.7 Effects on Operational Efficiency
The scientific consensus is that both across different methodologies and time periods PE ownership has a positive effect on management practices and opera-tional performance within the portfolio companies. The ratio of cash flow to sales has been seen to increase on average 40% when companies become PE owned. The ratio of operating income to sales has been seen to increase by 10-20%. Studies which have looked at oper-ational efficiency together with productivity have seen substantial increases in both. Something worth noting is that a large extent of the recent data on changes in the operational efficiency of PE owned companies since the 1980s has been gathered from Europe since data on pri-vate companies is more easily available there (Kaplan and Strömberg2009).
One particularly interesting study looked at the im-pact of PE investment on aggregate growth and cycli-cality in industries. This was assessed by analysing the relationship between the presence of PE investments and the growth rates of productivity, employment, and capital formation across 20 industries in 26 major na-tions between 1991 and 2007. It found that PE activity in an industry is associated with higher growth in both employment and productivity. Furthermore, it could
not find that PE presence increased the cyclicality of the industry significantly, which might have been expected as a result of higher leverage (Bernstein et al.2017).
The effect generally observed in terms of change in EBITDA margins as a measure of operational efficiency is that EBITDA margins increase by 3-4%. The motiva-tion is that this follows from a combinamotiva-tion of gover-nance, financial and operational measures taken within the firms (Grubb and Jonsson2007)(Viral V. Acharya and Kehoe2013).
2.8 Exit strategy
The exit is an important part of any private equity firm since this is when the return is realized. The most com-mon way to exit is the sale of the portfolio company to a strategic buyer which happens 38% of the time, the second most common is to sell to another private equity firm which happens 24% of the time, the route of an IPO is less common and only stands for 14% of all exits. Because of the high debt levels, one might expect a significant number of transactions to end in bankruptcy. However, the annual default rate for pri-vate equity owned companies is found to be 1.2% which is lower than the general average default rate of 1.6% reported by Moody’s (Kaplan and Strömberg2009).
2.9 Fund returns
The question of whether private equity firms earn supe-rior returns to their investors depend on several factors. First, private equity firms generally pay a premium when buying out companies, as a result the sellers should cap-ture a significant chunk of the value created by private equity firms. Secondly, private equity funds have high fees which reduce the return to investors in the fund (Kaplan and Strömberg2009).
(Kaplan and Schoar2005) found that LBO fund re-turns net of fees is slightly lower than those of the S&P 500. However, gross of fees the returns exceed those of the stock index.
Phalippou and Gottschalg (2009) used an improved and extended version of the dataset used by Kaplan and Schoar. Their study showed that net of fees PE funds underperform the S&P 500 by as much as -3.83% per year. However, gross of fees these funds outperform the S&P 500 by 2.96%.
of the funds they invest in and the effect they have on returns (Phalippou2009).
However, Harris, Jenkinson, and Kaplan (n.d.) found several faults with the datasets which had been used before. By analysing three new commercial datasets it consistently found that PE funds outperformed the S&P 500 by 3% per year or approximately 27% during the lifetime of the fund net of fees. Thus, painting a more positive view of the investor return from PE funds.
3
Mathematical background
3.1 Variables
The variables used in a regression model can be ei-ther numerical or categorical. Numerical regressor vari-ables have a well-defined scale of measurement and are the most commonly used type in regression mod-els (Montgomery, Peck, and Vining2012). When con-sidering certain properties that are relevant for the re-gression model, but not easily quantifiable, then it is necessary to use a categorical variable. This can be achieved by introducing a so-called dummy variable into the model, constructed by taking the value 1 if a given observation has a certain property, and value 0 if it is lacking it.
3.2 Multiple linear regression
Linear regression is an approach to model the relation-ship between parameters. A simple linear regression model, consisting of one response variable and one re-gressor, can be expressed on the form
yi = β0+ β1xi + i
where yi is the so called response variable and xi is
called a regressor variable. β1is a coefficient that in-dicates the proportionality between the two variables, while β0is the intercept of the model. i is the
error-term of the model. It is assumed that the error error-term should be normally distributed with mean zero, or
∈ N (0, σ2)
where the variance σ2is constant. With the ordinary least squares approach, the coefficients are determined by solving the minimization problem
mi n
n
Õ
i=1
(yi − ˆβ0− ˆβ1xi)2
This corresponds to minimizing the square distance of the residual.
If the response variable depends on several regres-sor variables, then the simple linear regression model can be extended to a multiple linear regression model. In this case, the model is on the form
y= βX +
where y = (y1, y2...yn), X = (x1, x2, ...xn), β =
(β0, β1...βn), = (0, 1, ...n). Similarly to the
sim-ple regression model, the parameters are obtained by solving the minimization problem
mi n (y − Xβ )0(y − Xβ )
(Montgomery, Peck, and Vining2012). As the problem is on quadratic form, solutions are obtained by derivation
∂
∂β(y − Xβ )
0
(y − Xβ )= 0
which in turn yields parameters as β= (X0X)−1X0y
Note that this assumes that the matrix X0Xis invert-ible, which in turn requires the assumption of no multi-collinearity to hold. The subject will be more thoroughly discussed later in this chapter.
3.3 Assumptions
The linear regression analysis approach is based on four assumptions that are the following:
1. Linearity and additivity 2. Statistical independence
3. Homoscedasticity (Constant variance) 4. Normality
The first assumption is crucial, due to that the model will otherwise perform poorly with high errors and un-reliable results. Lack of statistical independence is a sign that the model is not fully refined and that there is room for improvement. Violation of homoscedasticity results in difficulties to construct accurate confidence intervals, as they will be too narrow in some areas and too wide in others. Lastly, the normality assumption is crucial for evaluating whether the model coefficients significantly differ from zero (Nau2014).
3.4 Residual analysis
A residual is defined as
ei = yi − ˆyi, i = 1, 2, ..., n
where yiis an observed value and ˆyi is the value given
by the model (Montgomery, Peck, and Vining2012). The coefficient of determination, R2, defined as
R2 = S SR S ST
where S ST = S SR + S SR E S, S SR = n
Í
i=1(ˆyi − ¯yi) 2and
S SR E S = n
Í
i=1
(yi−ˆyi), and is a measure of the variability
in the response variable y that remains after consider-ing the impact from the regressor variables X. A value on R2that is close to 1 implies that the model has a high predictive accuracy, while a value close to 0 indicates that the model should be further refined.
A thorough analysis of the residual is essential to evaluate and refine a regression model. There are sev-eral methods and measures one can use when analysing the residual. An initial approach is often to perform a visual inspection by plotting the residual. This allows for a quick way to identify potential inadequacies in the dataset or the model. To simplify the analysis, the residual can be normalized by converting it to either a standardized or studentized residual. The studentized residual is defined as
ri =
ei
M SR es(1 − hi i)
where M SR esis an unbiased estimate of the variance
σ2and hi iis the corresponding diagonal element of the
hat matrix H. If the initial assumptions hold, then one can show that ri ∈ N (0, 1). By normalizing the residual,
one can identify potential outliers in the data. Points for which ri > 3 should be examined as potential
out-liers as99.9% of points should be concentrated within 3 standard deviations from the mean 0.
The prevalence of influence points can have a se-vere impact on a model. A commonly used measure for identifying influential points is Cook’s distance, defined as Di = ri2 p hi i (1 − hi i, i = 1, 2...n
where ri2is the square of the studentized residual and
hi i
(1−hi i)is a component that measures a points leverage
(Montgomery, Peck, and Vining2012). The measure is highly useful due to it reflecting how much a certain point impacts the model. Which values to examine is highly debated, but in his initial paper, Cook suggested that one should use F0.50(p, n − p) as a cut off value for influential points (Cook1977).
3.5 Multicolinearity
Multicollinearity is a problem that can seriously im-pact regression models precision and usefulness (Mont-gomery, Peck, and Vining2012). Multicollinearity occurs when there is a near-linear dependence between the regressors in a model. With prevalence of multicollinear-ity, the columns of the matrix X0Xbecome increasingly linearly dependant, resulting in det (X0X) → 0 and
(X0X)−1approaching a singularity. As a consequence, multicollinearity results in overestimated coefficients and high variance for the model (Carl. F. Mela.2002). Due to the dire consequences the problem can cause, one should carefully investigate the potential preva-lence of multicollinearity in a model.
A commonly used measure to identify multi-collinearity is the variance inflation factors, VIF, defined as
V I F = 1 1 − R2
j
where R2j is a coefficient for multiple determination that is computed by regressing a regressor variable xj on
the other regressor variables (Montgomery, Peck, and Vining2012). As a rule of thumb, regressors for which V I F > 10 implies a significant prevalence of multi-collinearity that must be managed through e.g variable selection or ridge regression to increase the model’s accuracy.
3.6 Model evaluation
The final stage of the model creation process is model evaluation. There is a multitude of different measures that can be used when evaluating which regressors to keep and if any final changes should be done to the model.
A T-test is used to determine the probability that a coefficient corresponding to a certain regressor dif-fers from 0. This is performed by testing the hypothesis H0 : ˆβi = 0. The hypothesis can be tested by forming
the statistic Z0= ˆβi σ2/Sx x ∈ tn−2 If it holds that |Z0|> tα /2,n−2
then the null hypothesis is false on significance level α and can be rejected. As σ2is rarely known, one can use the unbiased estimator σ2 ≈ M SR E S. Note that for this
test to be accurate, it is necessary that the normality assumption is satisfied.
variables should be included in the final model since a model based on only the most important regressors lowers the risk of imprecise results. Akaikes information criterium, AI C , is defined as
AI C = nl n(S SR E S n )+ 2p
and chooses a model based on regressors that maxi-mizes the expected entropy of the model (Montgomery, Peck, and Vining2012). The final model will then be based on a few key variables. Mallows’s CpStatistic is
a criterion related to the mean square error of a fitted value and is defined as
Cp =
S SR es(p)
ˆσ2 − n+ 2p
The criterion is especially useful for determining whether a certain regressor variable contributes to cre-ating a bias in the equation. It is desirable to have small values for Cp.
Commonly used algorithms for variable selection are three types of stepwise-type algorithms:
1. Forward selection 2. Backward elimination 3. Stepwise regression
Forward selection starts off with no regressors in the model, except for the intercept. It then seeks the op-timal regressor variable to include in the model, which is the regressor with the largest simple correlation with the response. The process is then repeated until the F statistic of the model exceeds a certain value.
Backward elimination works in a similar way to the forward selection. The main difference is that in the backwards elimination the model starts with a model based on all regressor variables and then proceeds to eliminate the variables that contributes the least to the F statistic.
Lastly, stepwise regression uses a dynamic program-ming approach by at each step identifying the combi-nation of variables that yields the highest F statistic.
4
Data
4.1 Description
The dataset which is used was generated from Orbis, a database containing information regarding private companies’ financial data. The private equity acquired companies that were studied encompasses a sample of transactions by 30 major private equity firms between the years 2011-2015 in the Nordic market. The result was a sample of 70 deals, out of which 56 companies had financial data of sufficient quality to be analysed.
The reference sample composed by randomly se-lecting 30 000 private companies in the Nordics which had not been acquired during the time period. The companies where chosen such that they were of similar size to the acquired companies. Relevant data which was collected relating the companies was revenue, EBIT, net current assets, fixed assets and full-time employees over a 5-year period. These datapoints were then used to form the regressor variables. See (Table1) for further descriptive statistics of the dataset.
4.2 Data management
The dataset had flaws that had to be addressed in or-der to perform the regression. Even though nearly all companies within the Orbis database showed accurate revenue and asset figures, the quality of EBIT and full-time employee reporting varied substantially. Since the focus of this analysis relates to the EBIT margin devel-opment over time, a substantial number of companies with low data quality had to be excluded both from the dataset of acquired companies and respectively from the reference dataset. However, no indication was seen that this data management resulted in a significant bias. Adjustments in the reference data were made for com-panies with incomplete data due to being to recently incorporated or acting as a holding company for sub-sidiaries. A major issue regarding the private equity companies was the restructuring of companies with ownership changes and changes in reporting standards, leading to a quite high number of companies needing to be removed.
Table 1.Descriptive Statistics of the Dataset. Source: Orbis
Statistic Non-acquired private company Acquired Company
Size of dataset 14 386 56 Size of Dataset Non-Adjusted 17989 70
Median Yearly Revenue 62 886th SEK 67 302 th SEK Median Number of Employees 40.5 34
Median EBIT Margin -0.3% 0.12% Number of PE firms N/A 30
Time span N/A Acquired 2011-15
5
Methodology
5.1 Model creation
To test the research question multiple linear regression was applied to the dataset. The response variable was defined as the difference between the average EBIT mar-gin during the year prior to being acquired and the av-erage margin from the acquisition year and three years forward.
yr es p = ¯y3−5− ¯y1−2
Regressor variables based on financial information for the model were defined as average yearly percent-age growth in revenue (xrev), fixed assets(xfix), net
cur-rent assets (xnca)and full time employees (xfte) during
the last three years of the relevant time span. Another regressor was the percentage increase in EBIT margin the year prior to the acquisition (xpre). Thus, capturing
the momentum effect of how previous development af-fect future development. Moreover, one regressor was defined as the absolute company size in terms of rev-enue (xtrev). Lastly, a categorical regressor variable was
defined as whether the company had been acquired or not. This was done by introducing a dummy-variable as
xacq =
(1, if company has been acquired 0, if company has not been acquired
Thus capturing the difference in EBIT margin develop-ment between PE owned companies and non PE owned private companies.
A multiple linear regression model was created on the form
yr es p = β0+ β1xrev+ β2xfix+ β3xnca+ β4xfte+
+β5xpre+ β6xtrev+ β7xacq
The parameters to the model where computed with OLS regression using the built in function lm() in R.
During the model creation it was found that com-panies which had negative initial EBIT margins differed significantly in their development compared to com-panies which had positive initial EBIT margins. Some companies yielded a change in EBIT margin well over 100%, which is theoretically impossible if the company has a positive EBIT margin. As a result, these datasets were separated into two different multivariable linear regression models. Thus, separating healthy compa-nies with initial positive EBIT margin from turn-around cases with initial negative EBIT margins.
5.2 Outlier management
Outliers were managed primarily by analysing the stu-dentized residuals and COOK’s distance. Points i where either ri > 3 or Di > F0.5(p, n − p) were examined
and removed from the data sample. A more precise approach would have been to look at each of these in-stances individually to determine if they actually are an outlier. Since individually examining a large number of companies, a sample of 30 companies were analysed to identify the cause for the unusual results. A significant number of these outliers had in common that they were in fact a holding company which affect reported results.
Furthermore, adjustments were made for extreme outliers in specific values. As a result, boundaries on FTE % growth, revenue % growth and EBIT % growth were put in place in order to adjust for unreasonable growth, for instance companies with an average 1000 % revenue growth year on year which create unnecessary noise in our model.
5.3 Multicollinearity
multi-collinearity was the relationship between revenue and FTE, which is expected due to companies needing to employ more people to expand. However, there was no need for adjustments as the highest value for the VIF was 1.577 for revenue growth. No regressors where removed due to 1.577 being significantly lower than the cut off value of 10 for VIF. See (Table4) in the appendix for more detailed results.
5.4 Variable selection
Forward selection, backward elimination and stepwise regression where performed to decide which regres-sors to use in the final model. After performing variable selection tests, the conclusion was that most regres-sors contributed significantly to the accuracy and per-formance of the model. However, the absolute size of the company in terms of revenue had low significance and no meaningful contribution to the accuracy of the model. Thus, a decision was made to exclude this vari-able from the model. The decision was made since both forward selection and backward elimination suggested the removal of the regressor. See (Table5) for forward variable selection for the positive EBIT model and (Ta-ble6) for forward variable selection for the negative EBIT model.
5.5 Management Theory and Model Choice
The choice of methodology is motivated both in man-agement theory and accounting theory. EBIT margin is a suitable metric of operational efficiency since it mea-sures the profitability of the operations. A potential is-sue with using EBIT margin is the potential impact from accounting changes that can impact the depreciation and amortization rate. Therefore, it could be advanta-geous to use EBITDA or EBITA as a measure of opera-tional efficiency. However, neither EBITDA or EBITA are recognized measure in the GAAP accounting standards, which would have resulted in a significant reduction in the quality and size of the dataset. To account for the potential yearly fluctuations, the response variable was based on average margins over the time period.
Regressor variables where chosen in order to maxi-mize the accuracy of the model. The momentum vari-able was included because it captures any ongoing projects and efforts to improve the EBIT-margin and can therefore be seen as a regressor based on previous management competence. Revenue growth indicates that the company is currently expanding, or the
under-lying market is growing, which logically could have an impact on the EBIT-margin. Tangible fixed assets and FTE are investments in the future of the company, and it is therefore of interest to see whether these invest-ments are made with similar profitability level as before. Net current assets was examined as a potential way to increase the operational margin could be by improv-ing the service level through e.g. increased inventory stocks. An increase in this regressor also has a direct link to other profitability measures such as return on work-ing capital. Lastly, the acquisition variable was included in order to capture the direct difference in operational efficiency as a result of the company being owned by a private equity firm. Consequently, the acquisition re-gressor is the most interesting factor in terms of the research question.
6
Results
6.1 Positive EBIT Model
The first model which results will be presented is the regression model based on companies with an initial positive EBIT margin. The model was deemed of good quality with a R2 = 0.6408. The regressor variables had an overall high significance level with low p-values. The studentized residual showed a satisfactory distributed relationship which is shown in (Figure2). The QQ-plot showed a linear relationship and is thus in line with the assumption of normally distributed residuals. The QQ-plot for both models is shown in (Figure1).
The results showed that being acquired by a PE company resulted in a 1.13 % reduction in EBIT margin compared to the reference companies over the studied time period. Furthermore, the regressor with the high-est explanatory power was the momentum variable, which was expected. Each additional % unit of EBIT margin the company increased during the beginning of the time period produced 0.7% units in extra future annual average EBIT growth per year during the rest of the period. Changes in revenue gave slightly positive re-sults on EBIT margin growth with a coefficient of 0.0131. Changes in the number of FTE gave a slight negative effect on EBIT margin growth, however the effect does not have a substantial impact on the prediction of mar-gin growth. The significance of the regressor variables where high with a p − v al ue <0.01 for all variables in the final model. For full results for the first model see (Table2).
Table 2.Results from positive EBIT model. Model accuracy: R2= 0.6408 , Model significance p-value < 2.2 ∗ 10−16
Regressor Estimate Std. Error t value Pr(<|t|)
Intercept -0.0036178 0.0002550 -14.188 < 2e-16 Acquisition -0.0113777 0.0041635 -2.733 0.00629 Revenue Growth 0.0131273 0.0008051 16.304 < 2e-16 Tangible fixed assets -0.0008003 0.0002629 -3.044 0.00234 Full Time Employees -0.0153092 0.0008219 -18.627 < 2e-16 Net Current Assets 0.0080790 0.0003568 22.642 < 2e-16 EBIT Margin Momentum 0.7226664 0.0051141 141.310 < 2e-16
6.2 Turn-around Model
The second model which results is to be analysed is based on companies which had initial negative EBIT margin during the first year of the time period. This model had even higher accuracy than the positive EBIT margin model with R2 = 0.8777. The QQ-plot and residuals where satisfactory and indicated normally dis-tributed error terms. However, the plots where less reassuring compared to those of the positive model.
In this case a PE acquisition had a positive impact on future average yearly % unit change in EBIT margin at the magnitude of 1.99%. However, the p-value
un-fortunately indicates a relatively low significance for the regressor. Once again changes in FTE and revenue produced highly significant effects, although small val-ues on the coefficients indicating a low impact on the response variable. The most significant driver of EBIT-margin growth was the momentum regressor. For turn-around cases the effect was even larger than was ob-served for companies with initial positive EBIT mar-gins. Each additional % unit of EBIT margin the com-pany had during the beginning of the time period pro-duced 0.9499% units in extra future annual average EBIT growth per year during the rest of the period. For full results of the negative EBIT model see (Table3)
Table 3.Results from negative initial EBIT model. Model accuracy: R2= 0.8777 , Model significance p-value < 2.2 ∗ 10−16
Regressor Estimate Std. Error t value Pr(<|t|)
Intercept 0.008159 0.001153 7.078 2.02e-12 Acquisition 0.019894 0.012367 1.974 0.0486 Revenue Growth 0.053636 0.003263 16.437 < 2e-16 Tangible Fixed Assets -0.002450 0.001212 -2.021 0.0435 Full Time Employees -0.030638 0.003264 -9.388 < 2e-16 Net current assets 0.001490 0.001332 1.118 0.2636 EBIT Margin Momentum 0.949931 0.008269 114.876 < 2e-16
7
Discussion
The results indicate a negative impact on EBIT over a time period of three years after being acquired by a pri-vate equity firm if the company had positive margins before the acquisition. On the other hand, the results indicate a positive impact on EBIT margin development if the company had a negative margin before the acqui-sition.
The results from the turn-around model are con-sistent with the general consensus that PE ownership
increase operational efficiency of their portfolio compa-nies. However, the results from the positive EBIT model differ from expectations and previous results.
complete financial data for all transactions would have been necessary. However, the belief is that the sample is large enough to give a representative, significant and useful answer to the research question.
It is premature to conclude that this study disproves the consensus that PE ownership has positive impact operational performance. However, since nearly all pre-vious research has been made on datasets consisting of PE exits made before 2005, it is not unreasonable that the nature of the industry has shifted substantially since then. Additionally, previous studies have not made the separation between healthy companies and turn-around companies but studied the industry as a whole. If this hypothesis holds as true, then it could indicate that private equity firms in the Nordic market are cur-rently more focused on increasing returns in healthy companies through multiple arbitrage and financial engineering. Such a choice of ownership strategy is attractive as it is less resource consuming compared to increasing operational efficiency. On the contrary, it is necessary to make major operational changes in turn-around companies due to it losing money when acquired which limits potential positive effects from financial engineering.
8
Areas for future research
Future research based on data of higher quality is sug-gested in order to confirm these findings. This study
only considered a three-year period after an acquisi-tion. An extension of the study would be to look at the entire ownership period as this data becomes available for the companies considered in this article.
Secondly, few researchers have differentiated be-tween turn-around PE investments and profitable growth cases. We believe that there is future research potential in make this distinction in order to further highlight the nature of how and when PE ownership increase the value of their portfolio firms.
9
Conclusions
This article has investigated the effect that private eq-uity ownership has had on operational efficiency in Nordic companies during the 2010s. The conclusion is that private equity ownership had a significant im-pact on EBIT margin of -1.14% if the company had a positive EBIT margin before the acquisition. On the other hand an acquisition had a 1.99% positive impact on EBIT margin development if the company had nega-tive EBIT margin before the acquisition. The conclusion is thus that private equity companies contributes to increased operational efficiency in turn-around acqui-sitions but decreases operational efficiency in already healthy companies.
References
Argentum. 2018. “The state of Nordic private equity 2018.”https://argentum.no/wp-content/uploads/sites/73/2019/04/Argentum_ The-state-of-Nordic-private-equity-2018_digital.pdf.
Bain. 2019. “Global Private Equity Report 2019.”https://www.bain.com/contentassets/875a49e26e9c4775942ec5b86084df0a/ bain_report_private_equity_report_2019.pdf.
Bernstein, S., J. Lerner, M. Sørensen, and P. Strömberg. 2017. “Private Equity and Industry Performance.” Management Science 63 (4): 1198–1213. doi:10.3386/w15632.
Carl. F. Mela., P. K. K. 2002. “The impact of collinearity on regression analysis: the asymmetric effect of negative and positive correlations.” Applied Economics 34.
Cook, R. D. 1977. “Detection of influential observations in linear regression.” Technometrics 19.
Davis, S. J., J. Haltiwanger, K. Handley, R. Jarmin, J. Lerner, and J. Miranda. 2014. “Private Equity, Jobs, and Productivity.” American Economic Review 104 (12): 3956–90. doi:10.1257/aer.104.12.3956.
Ehrs. 2020. “Den svenska private equity-industrin är världsledande.” Realtid. https://www.realtid.se/karriar/den-svenska-private-equity-industrin-ar-varldsledande.
Froud, J., and K. Williams. 2007. “Private Equity and the Culture of Value Extraction.” New Political Economy 12 (3): 405–420. doi:10.1080/13563460701485656.
Grubb, M., and S. Jonsson. 2007. “The Operating Impact of Buyouts in Sweden - A Study of Value Creation.” In Master’s Thesis in Finance Stockholm School of Economics.
Harris, R. S., T. Jenkinson, and S. N. Kaplan. n.d. “Private Equity Performance: What Do We Know?” The Journal of Finance. doi:https://ssrn.com/abstract=1932316.
Ivashina, V., and A. Kovner. 2011. “The Private Equity Advantage: Leveraged Buyout Firms and Relationship Banking.” The Review of Financial Studies 24 (7): 2462–2498. doi:https://www.jstor.org/stable/20869312.
Kaplan, S. N., and A. Schoar. 2005. “Private Equity Performance: Returns, Persistence, and Capital Flows.” The Journal of Finance 60 (4): 1791–1823. doi:https://doi.org/10.1111/j.1540-6261.2005.00780.x.
Kaplan, S. N., and P. Strömberg. 2009. “Leveraged Buyouts and Private Equity.” Journal of Economic Perspectives 23 (1): 121–146. doi:10.1257/jep.23.1.121.
Lerner, J., M. Sorensen, and P. Strömberg. 2011. “Private Equity and Long-Run Investment: The Case of Innovation.” The Journal of Finance 66 (2): 445–477. doi:https://doi.org/10.1111/j.1540-6261.2010.01639.x.
Leslie, P., and P. Oyer. 2009. “Managerial Incentives and Value Creation: Evidence from Private Equity.” In EFA 2009 Bergen Meeting Papers.https://ssrn.com/abstract=1341889.
Montgomery, D. C., E. A. Peck, and G. G. Vining. 2012. Introduction to Linear Regression Analysis fifth edition. John Wiley Sons, INC., Publication.
Nau, R. 2014. “Regression diagnostics: testing the assumptions of linear regression.”http://people.duke.edu/~rnau/testing.htm. Pedro, M., P. Fjellaker, and O. Camas. 2012. “Post IPO performance of private equity backed entities: Empirical Evidence of the
Nordic Market.” In Master thesis in Business Analysis and Performance Management International Business.
Phalippou, L. 2009. “Beware of Entering into Private Equity.” Journal of Economic Perspectives 23 (1): 147–166. doi:https://www. jstor.org/stable/27648298.
Phalippou, L., and O. Gottschalg. 2009. “The Performance of Private Equity Funds.” The Review of Financial Studies 22 (4): 1747–1776. doi:https://www.jstor.org/stable/30225708.
Robert Cressy, A. M., Frederico Munari. 2007. “Playing to their strengths? Evidence that specialization in the private equity industry confers competitive advantage.” Journal of Corporate Finance 13 (4): 647–669. doi:https://doi.org/10.1016/j.jcorpfin.2007.04. 007.
Rooney. 2020. “Private equity’s record $1.5 trillion cash pile comes with a new sof challenges.” CNBC.https://www.cnbc.com/2020/ 01/03/private-equitys-record-cash-pile-comes-with-a-new-set-of-challenges.html.
Spliid, R. 2013. “Is Nordic Private Equity Different?” The Journal of Private Equity 16 (2): 38–57. doi:https://www.jstor.org/stable/ 43503743.
10
Appendix
Table 4.Multicollinerarity table of VIF for both models
Regressor Positive EBIT Model Negative EBIT model
Acquisition 1.0068 1.0299 Revenue Growth 1.5775 1.5651 Tangible Fixed Assets 1.067 1.0593 Full Time Employees 1.4964 1.5081 Net Current Assets 1.0554 1.0122 Revenue Size 1.0440 1.0541 EBIT Margin Momentum 1.0102 1.0241
(a)Normal Probability Plot for positive EBIT model
(b)Normal Probability Plot for negative EBIT model Figure 1.Above is an illustration of the normality of error terms for the two models
Table 5.Forward selection algorithm results. Positive model
Step Regressor R-Square Adj. R-Square C(p) AIC RMSE
1 Pre 0.9981 0.9981 141115.9987 -33451.4752 0.0760 2 Rev 0.9982 0.9982 139426.9411 -33607.3366 0.0756 3 FTE 0.9982 0.9982 137297.7031 -33806.7788 0.0751 4 Cur 0.9982 0.9982 136858.2138 -33846.8616 0.0749 5 Fix 0.9982 0.9982 136771.8444 -33853.2997 0.0749 6 Acq 0.9982 0.9982 136754.1398 -33853.1819 0.0749
Table 6.Forward selection algorithm results. Negative model
Step Regressor R-Square Adj. R-Square C(p) AIC RMSE
1 Mom 0.9981 0.9981 28840.4281 -33451.4752 0.0760 2 Rev 0.9982 0.9982 28371.9400 -33607.3366 0.0756 3 FTE 0.9982 0.9982 27780.9843 -33806.7788 0.0751 4 Cur 0.9982 0.9982 27660.1526 -33846.8616 0.0749 5 Fix 0.9982 0.9982 27637.5663 -33853.2997 0.0749 6 Acq 0.9982 0.9982 27634.0841 -33853.1819 0.0749
(a)Studentized Residuals for positive EBIT model
(b)Studentized residuals for negatie EBIT model
TRITA 2020:119
