«BI Norwegian School of Management Hand-in date: 01.09.2011 Thesis supervisor: Øyvind Bøhren Program: Master of Science in Business and Economics ...»
When the debt holders are paid their ﬁxed fees, the residual claim ﬂows to the equity holders. We believe the net eﬀect of leverage will be reduced growth on the long term, because of the debt overhang problem.
Overall, there are several agency theories and we believe they aﬀect growth.
We test for the ﬁrst and third agency problem in our model.
2.4 Diversiﬁcation loss High ownership concentration, as in family ﬁrms, can create a diversiﬁcation loss. The family owner foregoes the beneﬁts of diversiﬁcation and require a compensation for this in terms of higher return. The portfolio would be many times riskier with a high concentration in one company, compered to spreading the risk through several smaller investments in diﬀerent ﬁrms(Bohren, 2011).
If he is forgoing the beneﬁts of diversiﬁcation, he will require a compensation of higher return to maintain his high ownership concentration. This is congruent with Meulbroek (2001) who found a large deadweight loss in stock and option rewards for managers who have a large ownership fraction in a corporation.
The diversiﬁcation loss can be particularly risky in combination with high leverage. It is also in line with empirics that shows owner-managed ﬁrms to grow slower but perhaps be more proﬁtable (Coad, 2007).
Alternatively, since the owner has the power, he/she is able to push the ﬁrm in the direction of more diversiﬁcation. If the ﬁrm is diversiﬁed, the standard 2 Literature review 10 deviation is lower, and this dampens the growth rate.
Summing up, unlisted family ﬁrms might face a diversiﬁcation loss, and lower growth rates, but will still be proﬁtable. We will not test for diversiﬁcation in our model since the evidence is not clear on diversiﬁcation and growth, but we will incorporate the arguments in the inference of our result.
2.5 Short-termism Family management might be associated with a diﬀerent proﬁt horizon compared to other shareholders. One reason might be that the family considers their role as to preserve the ﬁrm in order to survive through the current and future generations. Therefore, the horizon of family ﬁrms might span for years and decades, alternately professional managers are often criticized for being “short-termed”. Two Norwegian examples are Løvenskiold (started in 1649)2 which was owned and controlled the 13th generation, and Ugland (started the business activity in 1773)3. The payback period of many positive NPV projects can be longer than the expected horizons of the professional and public ﬁrm managers (Villalonga and Amit, 2010). This might turn out to be advantageous for the family ﬁrms if they are able to utilize a possible long-term premium.
To conclude, family ﬁrm might be more long term in their project selection, and hence be more proﬁtable in the long term.
2.6 Control variables Industry We adjust for industry sectors because diﬀerent sectors grow at diﬀerent rates. Some might also grow diﬀerently during various business cycle stages. We will use dummy variables to capture the industry eﬀect of nine diﬀerent sectors that are available in the CCGR data base. A similar approach has been used by Hamelin (2009).
Firm size Size is a variable that we believe has some inﬂuence on the growth rate. A large compilation of literature exists concerning size and growth, but we want to test for the size eﬀect through using the lagged size as a control.
Size is a common variable to adjust for especially in growth studies. Since small ﬁrms often grow faster than larger ﬁrms, it is quite important to adjust for size. Nevertheless, Gibrats law claims that growth is independent of size, butEvans (1987) falsiﬁes this. We follow Evans’ approach, because Gibrats http://www.lovenskiold.no/konsern/historie Retrieved 08/08/2011 http://www.jjuc.no/main.php?group=592 Retrieved 08/08/2011 2 Literature review 11 law has not yet been formally tested in the CCGR database (Berzins, Bohren, and Rydland, 2008).
Age Older ﬁrms might be less capable to be able to adapt to changes in the industry. On that note, they might lack the entrepreneurial desire to take risks.
The employees in older ﬁrms might also become entrenched in their routines and resist change. Evans (1987) proposes that the age does have a negative impact on the ﬁrms’ growth rate.
Financial constraints The corporate governance literature has brought up several diﬀerent problems aﬀecting a ﬁrm when the availability of capital is restricted. Debt overhang occurs when ﬁrms with high long term debt wouldn’t be ﬁnanced even if they had a highly proﬁtable project (Myers, 1977). Another problem is the asset substitution where highly leveraged ﬁrms choose short term projects even when the long term project gives a higher NPV. We expect ﬁnancial constraint to have a negative impact on growth Financial performance A more proﬁtable ﬁrm will be more capable in ﬁnancing projects with a higher portion of internal capital. We will expect a positive ﬁnancial performance to have a positive impact on the growth of the ﬁrm.
Separation of control If the CEO position is occupied by a member of the dominant family owner, it is often assumed that there is no separation between ownership and control. Furthermore, a CEO from the dominant family might have a negative impact on the growth, because of the many issues connected to family ownership.
2.7 Discussion Family ﬁrms have special characteristics that aﬀect the value of the ﬁrm. But the growth picture is still unclear since listed family ﬁrms grow faster than listed non-family ﬁrms. On the other hand, unlisted French SMB family ﬁrms grows slower than non-family unlisted ﬁrms.
The reduction of separation of ownership and control is an argument for higher growth because it gives clarity on who is controlling the ﬁrm. The development of the theories after Jensen and Meckling (1976) is more ambiguous.
Incompetence is a factor that could have a very negative impact on investing in the right project. Ineﬃcient investment due to due to owners’ personal interests is another factor that might lead to suboptimal investment decisions.
The longer horizon inherent in family ﬁrms would drive up the growth because 3 Model and variables 12 they are not over-investing in the projects which would be referred to as “short term projects”. In our opinion, the key issue with family ﬁrms is their unwillingness to raise new funds together with the desire to diversify, both these attributes contribute to slowing a growth rate down compared to non-family ﬁrms. The data and methods that have been used throughout the recent years is much more powerful than before, but proof of any diﬀerent diversiﬁcation between family ﬁrms and non-family ﬁrms seems to be a bit weak, but it still indicates that there are some diﬀerences. Diversiﬁcation may be a key for the growth family ﬁrms produce, but we will not include any diversiﬁcation variable in our model mostly because we are convinced that there will be noise in the measurements which will give less reliable inferences. Moreover, we are more convinced that there is some value in adding ﬁrm age, size and ﬁnancial performance as controls in our model as many other studies have done the same(Coad, 2009).
3 Model and variables
The model is designed to absorb the eﬀect family ﬁrms have on growth. We separate the eﬀect of A1 and A3 by using proxy variables for these theoretical constructs. In this way we might be able to interpret some of the unexplained growth that emerges from the A2 problem, but we cannot state that A2 is explaining all the residual growth yet to be explained, as there might be several other omitted variables.
Our empirical model can be summarized by the following econometrical
Growth We use revenue as a growth parameter. This is a common growth parameter used in corporate governance papers like (Schulze, Lubatkin, Dino, and Buchholtz, 2001) and in several articles about ﬁrm dynamics such as Huynh and Petrunia (2010).
Age We measure age by the year since the ﬁrm was founded. The Age is believed to follow exponential distribution (Coad and Tamvada, 2008). Therefore, we adjust our age-variable by squaring it to account for the non linear eﬀect.
Industry We will adjust for the growth in the diﬀerent sectors by using dummy variables for each sector. The sectors are classiﬁed by the NAIC coding making nine diﬀerent dummy variables. Adjusting for the industry average is a common practice and is done in the related study by Villalonga and Amit (2010) and Hamelin (2009).
Financial performance Our proxy for ﬁnancial performance is asset (ROA).
A ﬁrm with a higher proﬁtability has an easier job funding their investment projects. The return on asset is measured by the net income divided by the asset. ROA is a commonly used performance indicator in the literature as a measure of proﬁtability, see for instance Villalonga and Amit (2010).
Financial constraint We test for the problems mentioned in section 2.3 on page 9, because we think ﬁnancial constraints will aﬀect the family ﬁrms in our sample. We use debt/asset as our empirical proxy which is similar to the approach Villalonga and Amit (2010) have used. We expect ﬁnancial constraint to negatively aﬀect growth.
Family ﬁrm Our family dummy is created out of ultimate ownership data that matches family relationship by blood and marriage. Contrary to the thresholds mentioned in subsection 2.1 on page 6, we use a stricter threshold that requires 50% ownership held by the largest family in order to be deﬁned a family ﬁrm. Most of the unlisted ﬁrms in our sample have fewer owners than listed ﬁrms, which, we believe is an argument for a strict threshold. Conversely, if we had used a threshold of 20%, most of the ﬁrms would be family ﬁrms.
The CCGR database covers all Norwegian ﬁrms with limited liability in our sample throughout the period 1994-2009, and contains ﬁfteen years of accounting data. In Addition, nine years of governance data is included in the sample (2000-2009) (Berzins, Bohren, and Rydland, 2008). Relevant data are gathered from the database. The CCGR database will provide relevant data for our proxy variables in the time span 2000 to 2009 because ownership concentration on corporate governance data. We have used several ﬁlters on the dataset in order to produce a sample that is suitable and represent the ﬁrms we are interested in. The dataset is very large, consisting of many ﬁrms with either
poor accounting quality or little activity. The following ﬁlters are applied:
1. All the non-limited companies are removed.
2. All companies that have observations below 15 mill NOK in operational revenue are ﬁltered away.
3. All parent companies are eliminated from the sample.
4. All ﬁnancial companies are excluded from the sample.
5. All ﬁrms with no ownership data are excluded.
6. All the listed companies are removed.
7. All the ﬁrms with no asset are excluded from the sample.
8. All the ﬁrms with no employees are removed.
We ﬁltered out all micro-ﬁrms that had below 15 million in revenue as it is the threshold that the European Commission operates with, in their deﬁnition of a micro ﬁrm (Commission, 2003). We ﬁltered out micro-ﬁrms because they have extreme growth rates and make excessive amounts of noise. Also, it makes this study more comparable with similar studies of SMB ﬁrms in Europe. We also believe that SMB ﬁrms are suitable for our research objectives where we investigate family governance eﬀects on the growth, partly because there might be more owners in a SMB ﬁrm, compared with the micro-ﬁrms. Furthermore, listed ﬁrms are deleted because we believe they should not be mixed with unlisted ﬁrms, as they have a wider and perhaps more liquid sources of ﬁnancing opportunities than most unlisted ﬁrms. The latter is relevant because of the suggested ﬁnancing constraint mentioned in 2.6 on page 11. Moreover, we try to avoid listed ﬁrms to prevent taking the the double impact of a ﬁrm we need to take away the parent companies. The ﬁnancial companies are taken out of the sample because of their special accounting regulations. This makes them hard to compare to the other companies.
5 Methodology 15
We want to analyze our dataset using the most data points possible, to answer our research question. Our methods have been used before, but there has recently been a huge development of analytical methods for panel data. The ways we use panel methods has some known and unknown biases, e.g. that it is hard to predict causality (see section 5.2).
5.1 Panel data Panel data is more informative than that of a time series since it gives more data points which are able to be analyzed. With panel data, we look at both the ﬁrms in our sample, and watch them over time as well. This increases our number of observations considerably. Second advantage with panel data is that, with panel data, we are better able to study the growth of family businesses since we are conducting our study over time on a set of cross-sections observations. On the other side of the token, there exist problems with panel analysis as well. Cross section might exhibit heteroscedasticity, and over time;
autocorrelation. In addition, panel data might feature correlation between the ﬁrms at the same point in time.