• Ingen resultater fundet

– The CCT marginal contribution factor for bank loan level, total

The ratio is a simply the marginal contributions factor’s percentage impact on the level in year T. It tells the total value that the overall loan level increases over 1 year conditional on the capital structure.

In appendix 9 the marginal contribution factors for equity, bonds and bank loans are calculated.

67

5.3.2 Introducing the results ad interpretation of these

Below you will see two graphs and one table:

(i) Graph on the marginal contribution factor of the cross time-term on B. (ii) Graph on the marginal contribution factor of the cross time-term on E.

(iii) Table on marginal contribution factor of the cross time-term on BL. Due to the nature of the values a table is in this case more informative than graphs.

In each graph a trend appears for each country. The trick is now, that the only country characteristic included in the graph is the different structures. GDP and other individual characteristics are captured in the original regression .Each graph change over time due to two things: 1. the marginal yearly changes of the CTT estimator changes and 2. the capital- ratio changes. The latter only changes very little, i.e. the countries kept their ranking relative stabile over the periods of interest (see appendix 10) and therefore we can focus on the marginal yearly change. Below the marginal contribution factors for bonds, equity and bank loans are presented. The three sub-chapters do not follow the exact same structure, as some of the point made on bonds is the same for equity and thus not worth repeating. Since the marginal contribution factor for bank loans is calculated in a different why that for bonds and equity, it must be addressed in a different why.

5.3.2.1 Bonds

The following sub-chapter first present the marginal contribution factors for each country form 2005-2009. It then seeks to explain the changes over time from a mathematical point of view. These explanations can appear heavy and irrelevant, but approaching them systematically ensures that interpretation of the underlying economic reasons for the changes in level is causally valid and that we do not miss a point.

Bank-based countries are the ones with the highest marginal contribution factor regardless of the year in question. When (B/TL)T decreases, so does lag(log(B/TL))T, i.e. becomes more negative. The more negative log(B/TL)T, the higher the marginal contribution factor. The main question is then why the CTT estimators are negative and countries with low bond-ratios have the highest marginal contribution factors.

68 Graph 5.5 – Marginal contribution factor for bond level

The graph demonstrates the marginal contribution factor, i.e. the change in level over the recent year conditional on the bond-ratio, lag(log(B/TL). The cross time-term of each year is multiplied with the bond-ratio in each country in the same year. The bond-ratio, lag(log(B/TL)T, C = average the capital structure over the year.

Explanation of the factors and why they approach each other in bad times

For all countries at every point in time the marginal contribution factor stayed over 1 (France being an exception in year 2009). Thus, the cross time-term contribution factor grew in every period for every country and so did the equity level.

However, the speed by which the contribution factor grew decreased throughout the whole period until end 2008, i.e. the marginal contribution factor approached 1. Parallel the marginal contribution factors for each country approach each other, because they are a product of the marginal yearly changeT (∆ CTT estimatorT) and bond-ratio (lag(log(B/TL)T). When these two variables are multiplied the marginal contribution factor will be largest for countries with low (lag(log(B/TL)T) as their value is more negative. The difference between the factor of two countries with different bond-ratios, depends on the value of the marginal yearly change (the more negative the larger the difference).

The fact that we use a new cross time-term estimator each year allow the regression to tell if the claimed relationship between low (B/TL)T and large % changes in level since 2003 did not exist in a specific year.

The marginal contribution factors show that it is the bank-based countries that experienced the largest changes in the trend in level over the years. That is, they are the countries with the highest increase in level (or lowest decrease in 2008), but at the same time the marginal contribution factor changes with most percentage for these countries, i.e. the change in level

0,95 1,00 1,05 1,10 1,15 1,20 1,25 1,30

2005 2006 2007 2008 2009

Factor change in bond level conditional of the capital structure

Marginal contribution factor for bond level

Norway Denmark Spain England France Germany US Finland

69 (one can call it the Return) is more volatile for bank-based. Remembering that the return distribution of portfolios is more or less normal distributed and the higher volatility the higher the expected return. It appears as if the bond level in bank-based countries is simply more volatile (in contrast to our expectations). To support this argument it would be natural to look at the graphs in chapter 5.1. However, these graphs are not controlled for country specific characteristics and thus it is misleading to conclude anything from them. Instead, we can look at the standard deviation in the marginal contribution factor and it confirms that the bank-based countries (and France) were most exposed, see table 5.10. The underlying reasoning for why such tendencies occur is reflected upon in the discussion.

Table 5.10 – Volatility in the marginal contribution factor for bond level

Country Standard deviation in the marginal contribution factor

US 0,022

England 0,032

Finland 0,033

Norway 0,040

France 0,041

Denmark 0,042

Germany 0,049

Spain 0,080

Source: Appendix 8

Mathematical explanation of the relationship between the bond level and cross time-term Appendix 10.1 shows that the % ∆ in B from 2003 to 2009 is negatively correlated with (B/TL)2009. The same pattern is obvious when comparing % ∆ in B2003-T with any other year T. Thus, the higher % ∆ in level compared with the 2003 level, the smaller (B/TL)T ratio in the present years. It means that countries like the US, whose NFCs heavily use bonds, probably experienced less increase in the level through the years than countries with a very low dependency of bonds. Thus, the correlation coefficients support the cross time-term estimators. One reason for this tendency could be that countries with low (B/TL)Texperienced a larger increase in the (B/TL)T , i.e. a change in capital structure in favour of bond financing, and therefore they also experienced a higher increase in the level. However, appendix 10.1 demonstrates that a positive correlation between % ∆(B/TL)2003-T and the (B/TL)T - ratio. Hence, the countries mostly depended on bonds, experienced the higher increase. This stands in contrast to the cross time-term estimators. Thus, we can exclude the possibility that the

70 bond level increased more in low (B/TL)T –ratio countries, because of a change in preferences toward bond financing.

Instead, countries with a low bond - ratio can also have experienced a larger increase in the bond level over time, because the whole economies have expanded more. This would require not only the bond level, but also the equity and bank loan level to increase more in these countries. Estimates of the correlation between (B/TL)Tand % ∆ E2003-T is negative, and so it the correlation between (B/TL)T and % ∆ BL2003-T (Appendix 10.2 and 10.3). Thus, the lower (B/TL)T, the more did the bond, equity and bank loan level increase percentagewise and the more percentage did the total level of external finance increase.

Therefore, bank-based countries apparently went through a larger increase in all types of finance from 2003 to 2009 and therefore the bond level increased more in these countries, even though the bond-ratio changed less. It must be added that correlation figures not necessarily are significant. However, they still prove the point that bank-based countries not only experienced a more positive change in the bond level due to an overall larger increase in liabilities of the NFCs, but that they did so even though the market-based countries parallelly experienced a higher increase in (B/TL)T. One could argue that the reason why market-based countries experienced a larger increase in the bond-ratio is simply because they did not go through as large an increase in equity and bank loan level as bank-based countries did.

One very important property of correlations must be mentioned: Negative correlations merely state that there is a tendency of one variable increasing as the other decreases. This is not contradicting to a general decrease in both variables. Thus; the level could in theory decrease for all countries through times, but we will then see that the decrease is smaller in bank-oriented countries.

5.3.2.2 Equity

The marginal contribution factors are to be read in the same manner as bond level. The yearly change in equity level follows a slightly different order than that of bonds, but interestingly three out of the four least solvent countries (measured for NFC’s), are also the ones with the lowest bond – ratios, i.e. they seem to prefer neither bonds nor equity especially high. This is Norway, Spain and Germany.

71 Graph 6.6 – Marginal contribution factor for equity level

The graph demonstrates the marginal contribution factor, i.e. the change in level over the recent year conditional on the equity-ratio, lag(log(E/TL). The cross time-term of each year is multiplied with the equity-ratio in each country in the same year. The equity-ratio, lag(log(E/TL)T, C = average the capital structure over the year.

The reader might be surprised over the moderate decrease in level in 2008, as everybody knows that market value of equity dropped severely in 2008-2009. It is important to remember that the marginal contribution factors only focus on changes in equity level that can be explained by the financial structure in the countries used in the paper.

The factor emphasises that the countries percentagewise mostly affected by marginal yearly level changes, are the least equity-based ones. The reason is exactly the same as for bonds. In appendix 10.2 the correlation between (E/TL)2009 and % ∆ E2003-2009 is negative and the same pattern is visible for all other years. There exists a negative correlation between the % ∆ B2003-2007

and % ∆ BL2003-2007 and the (E/TL)T- ratio in the same years. The same tendency can be found in 2008 and 2009. Therefore a certain level of evidence support that the bank-based countries with low (E/TL)T experienced a more positive change in total finance. However, the correlation coefficients supporting such evidence are weak and no more attention will therefore be put into these findings.

The volatility in the marginal yearly change is larger for equity than for bonds and so it turns negative in 2008. Comparing Table 5.10 and 5.11 provide evidence of volatility higher related to equity than to bonds. In addition Table 5.11 shows that 3 out of the four most volatile countries are bank-based.

0,85 0,95 1,05 1,15 1,25 1,35 1,45

2005 2006 2007 2008 2009

Equity

Marginal contribution factor

Norway Denmark Spain England France Germany US Finland

72 For the same reason the marginal contribution factors approach each other more than for bonds. Note that Graph 5.6 shows that the countries shifts order, so the bank-based countries in 2008 were the ones experiencing the largest drop (the contribution factor is less than 1) in level.

The mathematical reason is that the marginal yearly effect turns positive, but lag(log(E/TL)T) still is most negative for the countries with a low (E/TL)T. Multiplying the two variables resulted in lower marginal contribution factor for bank-based economies.

In the same manner the bank-based countries experience that largest increase in equity level through 2009 and as for bonds they turn out to be the countries with the most volatile level due to their structural characteristics. Interpretation will follow in the discussion.

5.3.2.3 Bank loans

The marginal contribution factors calculated with regards to bank loans are so close to each other that a graph presenting the results would be hard to read. Instead the marginal contribution factors are presented themselves:

Table 5.12 – Marginal contribution factors on bank loan level for each country

Country 2005 2006 2007 2008 2009

Germany 1,407 1,297 1,173 0,976 0,978

France 1,419 1,309 1,179 0,977 1,019

Denmark 1,422 1,320 1,190 0,976 0,979

Finland 1,428 1,334 1,192 0,977 0,981

Norway 1,431 1,313 1,182 0,976 0,978

England 1,439 1,306 1,172 0,976 0,978

Spain 1,485 1,324 1,179 0,976 0,978

US 1,529 1,351 1,190 0,976 0,979

Standard deviation

/average factor 2,83% 1,29% 0,64% 0,05% 1,43%

Source: Appendix 4.3 for the CTT estimators and Appendix 8 for calculation of the marginal contribution factors

The table present the marginal contribution factor calculated as Equation 5.6. The percentage standard deviation from the mean is calculated, as it highlights how close the values follow each other. Equation 5.5:

.

Table 5.11 – Volatility in the marginal contribution factor for equity level

Country Standard deviation in the marginal contribution factor

US 0.058

Finland 0.079

Denmark 0.093

England 0.101

Germany 0.102

Spain 0.126

Norway 0.128

France 0.190

Source: Appendix 8

73 Explanation of why the country exposure deviates as little as it does

Table 5.12 emphasis a crucial point: The % ∆ in the bank loan level from the previous period for each country varied with only 2.83 % to 0.05%. The ranking follows no system with regards to the capital structure. This might appear contradicting to the fact that cross time-terms exactly proved that a significant relationship between level change over time and (BL/TL) existed. Below I try to explain why the variation was so small between the different financial systems.

As the marginal yearly effect is the same for all countries in a given year, the lack of variance is mathematically explained by a constant relationship between the bank loan-ratio and the predicted level minus the country specific impact (See Equation 5.4-5.6), i.e.;

Constant =

.

Such constant is quite intuitively, as we have a constitutive variable, lag(BL/TL)T, which on one hand claim a constant linear relationship with the bank loan level if controlling for country specific impact and time. It that sense, the marginal contribution factor simply supports the constitutive variable in contrast the factors for bond and equity level. However, with regards to the contribution factor of bonds, the cross time-terms predicted the relationship between lag(B/TL)Tand B to be less positive the higher lag(B/TL)T. The same was true for equity. This is not the case for the cross time-term linked to bank loans: If the level is predicted to increase in 2005 relative to 2003 and the relationship between BL and lag(BL/TL)T

is truly constant over time, the absolute level must increase more for high lag(BL/TL)T

countries. Only so will the relationship stay constant. This is exactly what the cross time-term estimators for bank loans predict. The marginal contribution factor for bank-loans shows the interesting point; it provides no new information in addition to its constitutive variables, i.e.

time dummies and bank loan-ratio as long as they both are used to predict the level at the same time.

But, the constant spells out how much more the level must increase more for high lag(BL/TL)T countries than low lag(BL/TL)T to keep the relationship constant.

A very important point support the explanation above: Correlation estimates between lag(BL/TL)T and % ∆ BL from 2003 to T shows no convincing signs, but for what it is worth these signs are positive (see appendix 10.2). Therefore, there is no convincing evidence for why countries with certain capital structure characteristics should change their level

74 percentagewise more than others and thus there is no reason why the time dummies and the bank loan-ratio variables should change relationship when combined. If we should read anything into the positive signs of the correlation, it merely supports the constitutive variables and perhaps shows weak signs of a synergy between terms13. The graph on bank loan level in chapter 5.1 supports the correlation estimates: No distinct system appears.

The general trend in the marginal contribution factor

It is important to highlight that changes over time did happen, but the change was almost the same for all countries and their exposure to level changes was rather constant.

Compared with the % ∆ in level for equity and bonds, bank loan level increases significantly more in 2005 and 2006. Actually the level increases with around 40 % in 2005 and 30 % in 2006 (see Table 5.12). It does not decrease as much as equity in 2008, but on the other hand it continues to decrease in 2009. Does it mean that bank loans are less volatile over time? Table 5.13 provides the answers. First, there are no convincing signs of what system that experienced the highest volatility in the contribution factor for bank loans from 2005 to 2009.

However, comparing Table 5.10 to Table 5.11 and Table 5.13, the marginal contribution factor shows much stronger volatility for bank loans in general than for equity and bonds. This is unexpected and will be discussed later. However, the result is that countries for whom the lag(BL/TL)T

is high, will all things equal feel a higher volatility in the total level.

Bank loans’ overall exposure to the crisis, relative to that of bond and equity, plays a leading in whether the total level of external finance in bank-based or market-based countries is most exposed. The answer to which system that was overall most exposed to the crisis is provided below.

13 A synergetic interaction term is one where its two reference variables, the constitutive variables, both provide marginal impact on the dependent variable, but where there exist a positive spillover effect. An example could be that a child, who is bright, receives better grades and a child, who is working hard, receives better grades. But a child that possess both skills receive better grades than simply adding the effect of intelligence and work effort.

Table 5.13 – Volatility in the marginal contribution factor for bank loan level

Country Standard deviation in the marginal contribution factor

France 0,188

Germany 0,192

Denmark 0,200

Norway 0,202

England 0,203

Finland 0,204

Spain 0,221

US 0,240

Source: Appendix 8

75

5.3.4 Change on total external finance

The total % ∆ in external finance in year T is calculated in the following way: