The effect of quantitative easing on occupational pension funds

1

Bachelor Thesis

BSc. Economics and Business – Finance and Organization Field: Finance

The Effect of Quantitative Easing on Occupational Pension

Funds

David Bolscher

10498346

University of Amsterdam (UVA)

20th of June, 2018

Supervisor: Spyridon Terovitis

Abstract

This paper analyses the effect of the Quantitative Easing (QE) program from the European Central Bank (ECB) on pension funds. The low interest rate can endanger the financial stability of long-term savers, such as pension funds. Performing a regression on the pension funds of nine Eurozone countries over the years 2010-2016, the results show the opposite effect. The decrease in interest rate seems to increase the financial stability of pension funds. However, this result might also be attributed to the measures pension funds took after the announcement of the QE program.

2

Statement of Originality

This document is written by Student David Bolscher who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3

1. Introduction

After the financial crisis in 2007, governments struggled with low inflation which resulted from a contraction in the credit market. Because low levels of inflation can hamper economic recovery, the Federal Reserve (Fed) introduced an unconventional monetary measure. This measure, called Quantitative Easing (QE), aims to raise inflation by lowering the interest rate. After the Fed, other authorities such as the European Central Bank (ECB) also started a QE program. The program from the ECB consists of a monthly capital injection which reduces the interest rate and thereby stimulates investment (Claeys, Leandro, & Mandra, 2015, p. 2). Although the inflation level resulting from this program induces fast economic recovery, there might also be some downsides. Not only can monetary expansion lead to new asset bubbles because low interest rates increase the demand for equity. Low interest rates can also affect savers who do not want or are not allowed to take risks. Besides, small savers such as households, some large savers as pension funds and life insurers might be affected by low interest rates.

In a recent study from Boubaker, Gounopoulos, Nguyen and Paltalidis (2016) on the QE program in the US, it appeared that pension funds switched towards a riskier portfolio with more short-term investment due to low interest rate. By shifting from long-term government bonds to equity, pension funds could protect themselves against a funding gap caused by the QE program. However, pension funds in Europe have less possibilities that partake in risk taking behaviour due to more severe regulations. These regulations have led pension funds in Europe to act in the opposite way to pension funds in the US. The announcement from the ECB resulted in pension funds starting to gather long-term government bonds in order to prevent a duration mismatch in their portfolio (Domanski, Shin, & Sushko, 2017, p. 121). Such a difference in behaviour in portfolio rebalancing indicates a risk for the European pension funds. Namely, their inability to rebalance their portfolio raise problems for them as the pension funds might become underfunded. Therefore, this paper will examine whether the QE program from the ECB harms pension funds. The effect of a low interest rate on the cover ratio of pension funds will be analysed. This analysis gives useful insights into the effects of protracted low interest rates. Not only does it make clear to what extent pension funds might be harmed, but also it empirically tests the theory indicating that low interest rates affect long-term investors. As such, the research can be applicable to other industries, for example to life insurers. This research is also an addition to the current literature about risk taking behaviour of long term investors in times of protracted low interest rates.

4

To test the effect of the QE program on pension funds a GLS regression with random effects will be conducted on panel data. The data is retrieved from the European Insurance and Occupational Pension Funds Authority (EIOPA) and consists of yearly statistics from pension funds averaged by country. It includes nine Eurozone countries and the years 2010-2016. The main findings of this regression indicate that the QE program does not affect pension funds. The hypothesis that the interest rate has a positive effect on the cover ratio of pension funds is not confirmed because exactly the opposite effect has been found. This does not mean that pension funds are not at all affected by the QE program or that the QE program has a positive effect on pension funds. As Domansky, Shin and Sushko (2017) point out, pension funds started buying long-term government bonds after the ECB announced their QE program. This might have prevented their cover ratios to decline, but can have an adverse effect in the long run. Furthermore, the results show that pension funds invest more in risky assets when the interest rate declines. Portfolio rebalancing brings more risk and uncertainty into the pension funds.

In the next section the theory underlying the main hypothesis will be explained. It contains theory on QE programs as well as on the cover ratio of pension funds. From this the hypothesis will be derived. The third section will discuss the methodology. The research design, dataset and validity of the regression will be discussed. The fourth section will discuss the results and test the hypothesis. In the final section the main findings will be summarised and there will be given some limitations and some recommendations for future research.

5

2. Literature Review

Before analysing whether the QE program lets pension funds run into deficits, the current literature should be explained. The literature review will show how this paper relates to other research on unconventional monetary policies and its effects on pension funds. Furthermore, this section will show how a low interest rate might be harmful to savers, especially to pension funds. Firstly, unconventional monetary policy and its potential consequences in general will be explained. It will be made clear how such a program leads to low interest rate levels which can be harmful to long-term savers. Thereafter, some papers which analysed the effects of unconventional monetary policy will be quoted. I will make clear how this research relates to these papers. Finally, it will be shown what the consequences of changes in interest rates are for duration mismatches, which leads us to the hypothesis.

Unconventional monetary policies started to be used by central banks when lowering the interest rate was not sufficient anymore. According to Joyce, Milles, Scott and Vayanos (2012, p. F271), maintaining a low and stable inflation by setting the interest rate did not prevent the problem of asset bubbles. Therefore, a new tool for monetary policy should be introduced. Instead of setting a lower interest rate, an expansion (or contraction) of the balance sheet of central banks could be used to achieve a certain level of inflation and account for asset bubbles. There are a few channels through which unconventional monetary policy might influence the economy, of which two will be explained in this article.

The first channel through which the policy can influence the economy is the balance sheet. It is implied that differences in assets and in investors’ preferences can be used by the central bank to stimulate investments. This theory is based on the preferred-habitat theory and on the idea that assets are not perfect substitutes (Joyce et al., 2012, p. F278). Because short- and long-term assets are not perfect substitutes, central banks could influence one of the two. Without preferences everyone will trade their long-term assets for short-term assets when they buy long-term assets. However, because some investors such as pension funds and insurers prefer long-term assets, the price of long-term assets could increase and thereby lower the yield. In this way long-term loans become available at lower costs, and households could save money. Finally, this process could lead to a higher demand for goods and, as such, to a higher Gross Domestic Product (GDP). Another, maybe weaker, channel is the bank funding channel. As a result of buying government bonds, bank deposits would rise. The reserves of banks increase and they would become more willingly to expand their lending.

6

In the US the theory behind the balance sheet channel was not fully verified with empirical evidence from the effects of the unconventional monetary policy level of the FED. According to Boubaker et al. (2017, p. 36) pension funds shifted towards equity after the FED started to buy government bonds. The lower yield in term assets not only pushed long-term investors towards short-long-term assets, also regulations in discounting played an important role. Whereas in Europe pension funds are required to use long-term yields for discounting their liabilities, US pension funds can use their expected return on assets as discount rate (Andonov, Bauer, & Cremers, 2017, pp. 1-2). Because the expected return of equity increased as a result of the unconventional monetary policy, shifting towards equity allowed pension funds in the US to discount their liabilities using a higher discount rate. As a result, the market value of liabilities did not increase significantly just like their cover ratio. Boubaker et al. (2017) thus looked into the effect of unconventional monetary policy on pension funds. Their research showed there was an incentive to undertake more risk in their portfolios. Although this paper uses the theory from Boubaker et al. (2017), it departs from it in two different directions. On the one hand this paper will focus on the Eurozone instead of the US. Because of a difference in regulations this might increase the external validity. On the other hand, this paper will not look at the reaction of pension funds on unconventional monetary policies, but whether they are affected if they would not react by undertaking more risk. By controlling for the composition of their portfolio, it becomes clear what might have happened to pension funds if they did not switch to riskier investments.

Besides such a difference in regulations, the difference in pension scheme is also important for analysing the effects of unconventional monetary policy on pension funds. The most common pension schemes are defined-contribution (DC) schemes and defined-benefit (DB) schemes. Where DC schemes pay-out their benefits on a variable basis, keeping their contributions fixed, DB schemes pay-out their benefits on a fixed basis. Although it is only used as a last resort, DB schemes could raise their contribution to be able to pay-out the fixed benefits (Antolin & Yermo, 2011, p. 239). In principle DC schemes could not be affected by unconventional monetary policy, because their pay-outs always equal their benefits. This means that DC schemes always have a cover ratio of 100%, because the market value of their liabilities adjusts to the market value of their assets. In practice the receivers of DC pension schemes could be harmed, because they would receive lower pay-outs.

The cover ratio, also called funding ratio, is the market value of assets divided by the market value of liabilities. Because the cover ratio is an indication about whether a fund can meet its obligations in the future based on the expected income on assets, it is an important

7

indicator for financial health (Brucciol & Beetsma, 2010, p. 367). Also, it is the main target for regulations in the Eurozone. Not only do countries require a minimum funding rate above 100%, they also prescribe how the assets and liabilities should be valuated (Boon, Brière, & Rigot, 2018, p. 25). Based on the IAS 19 regulations, Eurozone countries must discount their assets and liabilities with the market interest rate. This change can make the funding ratio of pension funds more volatile. Beforehand a fixed rate was used for discounting, thereby making it easier for pension funds to keep it constant over time. Because of this, the cover ratio of pension funds in the Eurozone will react on changes in the interest rate.

Thus, changes in the interest rate effect the cover ratio of pension funds because it is used as a discount rate. However, the direction of such a change depends on the duration of assets versus the duration of liabilities. According to Domansky, Shin and Sushko (2017, p. 114) institutions with a longer maturity profile try to obtain financial stability by matching the duration of assets with the duration of liabilities. They argue that pension funds and life-insurers have a negative duration gap which means that their obligations last longer than their assets. A negative duration gap implies that a change in interest rates results in a larger change in liabilities than in assets. When the interest rate rises, the discounted value of liabilities decreases more than the discounted value in assets. This would result in an increase in the cover ratio. On the other hand, when the interest rate falls, which is a consequence of the QE program, the discounted value of liabilities increases more than the discounted value in assets. This would result in a decreased cover ratio. Domansky, Shin and Sushko (2017) found that after the announcement of the QE program, life-insurers wanted to increase the duration of their assets by buying long-term bonds. Because the ECB started doing the same, the bond yields kept falling and therefore again worsened the situation of long-term savers. Their research provides some relevant theory, namely that pension funds have a negative duration gap. However, it only guesses that the gathering of long-term bonds is going to have a negative effect on pension funds. This paper contributes to this guess because it tests whether pension funds were affected. So, whether gathering long-term bonds indeed had a negative effect on pension funds.

European pension funds must use the market interest rate in their valuation methods. Because of their negative duration gap, the QE program might harm pension funds. In the analysis the hypothesis that a lower interest rate leads to a lower cover ratio will be tested. This will only be done for DB pension schemes, because their cover ratio can rise and fall. This test is relevant for answering the question whether unconventional monetary policy affects pension funds or other long-term savers. Furthermore, it is an addition to the research of Boubaker et

8

al. (2017) because it uses Europe instead of the US and thereby it addresses pension funds under more heavily regulations. Besides that, it is an addition to the research of Domansky, Shin and Sushko (2017) because it indicates whether the QE program harmed pension funds despite them buying long-term bonds.

9

3. Method

To test the hypothesis a longitudinal regression will be conducted. This paragraph describes the design of the regression and will outline some important concerns about validity. First, it will be explained where the data comes from and why the dataset can be used for testing the hypothesis. Afterwards the time span of the data will be motivated. Furthermore, all variables of the regression will be explained and motivated. The correlations between those will be discussed. Finally, the research method will be explained and the regression formula will be given.

To examine the impact of the QE program on pension funds a longitudinal study on occupational pension funds from 9 Eurozone countries over the years 2010-2016 will be conducted. The data on occupational pension funds from these countries is gathered by the European Insurance and Occupational Pensions Authority (EIOPA). This institution gathers information from countries to analyse the prospects for the insurance and pension sector. One of the main advantages of this dataset is that it includes important variables and many countries. It includes the cover ratio and the kind of investments pension funds undertake as well as the number of retired versus active members. Moreover, the dataset consists of occupational pension funds for which the regulations are clearer than for other pension funds. This could be helpful by examining the validity of the research. Because the regulations are severe, there are less possibilities for pension funds to manipulate their cover ratio. Therefore, it is expected that effects on occupational pension funds are representative for the effects on the whole sector, even when these effects cannot be found at pension funds which face less severe regulations. Because of the long duration of the liabilities, occupational funds are an example of long-term savers. Their capital structure is the same as other long-term savers, such as life insurers. For this reason, the effect of the change in interest rate will also apply to other long-term savers. However, the dataset might lead to some problems in the regression. To begin with, the data is yearly. The effect of the interest rate on the cover ratio will be tested, but it might be that there is not enough variation in yearly data. Furthermore, countries independently computed their data which is reported to EIOPA. This could have led to accounting differences. Finally, a problem for this research can be that the dataset consists of averages from countries instead of data from individual pension funds. It cannot be made clear how the distribution within countries can influence the results.

The years 2010-2016 from the dataset will be used in the regression. This is partly due to missing values before the year 2010. Also, the time span does not fully include the effects

10

of the financial crisis. Because the economy was still not fully recovered from the crisis, additional control variables will be included, these will be discussed later. Another important reason for choosing this time span is that it includes three years before and three years after the implementation of the QE program. This makes it easier to test for a break in the regression. Ideally the timespan would cover 2017 as well, as it is expected that the effects of QE have remained present in 2017, but the data of this year has not been published yet. In addition to this, the amount of observations which can be used from the dataset is small. So future research could increase the reliability by including more years in the regression and by testing for the years where the effects of QE disappeared.

As the dataset and its time span have been discussed, the paper will continue by explaining the dependent and explanatory variables. The cover ratio FR is the market value of the assets divided by the market value of the liabilities. When the cover ratio is beneath 100%, the expected income on assets is lower than the expected obligations, which indicates that the fund is in a critical financial situation. Therefore, authorities of countries force pension funds to undertake measures when their cover ratio becomes too low. For that reason, the cover ratio is being seen as the most important indicator for financial health of pension funds (Brucciol & Beetsma, 2010, p. 367). This makes the cover ratio a good instrument to test whether pension funds are financially affected by the QE program. A cover ratio could only be used for Defined-Benefit (DB) pension schemes, because their liabilities remain the same after a change in the market value of their assets. Pensions with Defined-Contribution (DC) schemes their liabilities automatically decrease with the value of their assets, resulting in a constant cover ratio of 100%. Therefore, the regression only includes pension funds with DB schemes. There are 9 Eurozone countries with such schemes, only these countries will be included.

The explanatory variable which is expected to be positively correlated with the cover ratio is the interest rate. Because pension funds use the interest rate in discounting for determining the market value of their assets and liabilities, it must influence the cover ratio. The assets are expected to have a lower duration than the liabilities, therefore a lower interest rate leads to a larger increase in liabilities than in assets. This would result in a lower cover ratio. Additionally, the interest rate is a proxy of monetary policy. The market swap rate will be used as the interest rate, because pension funds are required to use this rate for discounting since 2006 (Bikker, Knaap, & Romp, 2014, p. 2001). The data is retrieved from the ECB.

According to Brucciol and Beetsma (2010, p. 375) the cover ratio can be influenced by three types of shocks, namely economic, financial and demographic shocks. Economic shocks, such as inflation or nominal income, can influence the value of the assets and liabilities of

11

pension funds. Because inflation would depend on the interest rate, it is left out of the regression formula. However, nominal income will be accounted for by including the GDP per country i and year j. Data on the GDP of countries will be retrieved from Eurostat and will be indexed with 2010 as base year. GDP is an important variable to control for the effects of the financial crisis in 2007 and for other country dependent economic measures.

Besides GDP, the return on equity also influences the cover ratio as higher return on equity results in an increase in market value of the assets. Return on equity will be used as an indicator for financial shocks (Brucciol & Beetsma, 2010, p. 375). Moreover, it will be used to control for the financial crisis. Because the pension funds are from Eurozone countries, data on the Euronext 100 index, consisting of the largest stocks in the Eurozone, will be used as equity returns in the regression. The variable E consists of returns on the Euronext 100 index per year j, which data comes from yahoo finance.

According to Brucciol and Beetsma (2010) the final factor which influences the cover ratio of pension funds is demographic shocks. For example, these shocks could be changes in expected survival age or changes in retirement age. The variation in these factors is expected to be low, because the survival and retirement age can remain the same over 6 years for a particular country. Therefore, in the regression there will be controlled for demographic shocks by including the ratio of the net contributions divided by the net benefits. The net contributions are the amount of money received from active members and the net benefits are the amount of money payed out to retired members. Therefore, the C/B ratio accounts for the current survival age, as a higher survival age results in a lower C/B ratio. A disadvantage is that the expected survival age does not influence the C/B ratio and is for this reason left out of the regression. Data on the C/B ratio per country i and year j is retrieved from the dataset on pension funds from EIOPA.

Table 1. Descriptive statistics variables

Variable Obs Mean Std. Dev. Min Max

coverratio 63 1.077682 .1207225 .5901724 1.296401 interestrate 63 .0173014 .0106112 -.00057 .03171 r 63 1.833081 2.197804 .2044195 9.451343 gdpindex 63 103.4079 9.199059 93.2 149.7 n100 63 .0533749 .1008113 -.1434174 .1521778 cb 63 1.10877 .8737059 0 5.449072

12

Finally, the descriptive statistics of the variables will be briefly explained. For the cover ratio, the mean of 107.76% is expected because it means that pension funds can meet their obligations on average. However, the minimum value is low with a cover ratio of 59% found in Germany in 2012. This might be an unlikely value because one year later the cover ratio in Germany almost doubled to 115.72%. This could indicate that the pension funds received financial aid from the government. However, there cannot be found a clear reason for this increase. For this reason, it is expected to be a mistake and therefore it has been deleted from the dataset. No unexpected values can be found for the interest rate. With a mean of 1.73%, a minimum of -0.06% and a maximum of 3.17% the interest rate suits the expectations for the Eurozone in this timespan.

The descriptive statistics of the control variables show some unexpected values. First, the risk-taking behaviour has a wide range. Especially the maximum value is high with 9.451, because the values of the year before and after this year are a lot lower, beneath 1, it is expected that the value is wrong. Therefore, it will be erased from the dataset. The GDP index increased during those years on average for all countries. Also, the lowest value is close to 93 and the highest value is 150 which indicates an increase in GDP. The increase was expected as a result from the financial crisis in 2007. The returns of the n100 seems to be normal looking at the average and at the range. The cb variable contained some strange values, namely those with a ratio of zero. Because this would mean that the contributions were close to zero or small compared to the benefits, this would not be realistic. Therefore, all cb values smaller than 0.2 have been erased, 9 values in total.

Table 2. Correlation Matrix

coverratio interestrate r gdpindex n100 cb coverratio 1.0000 interestrate -0.2091 1.0000 r 0.6384 -0.0036 1.0000 gdpindex 0.3433 -0.3707 0.1295 1.0000 n100 0.2719 -0.2662 0.0336 -0.1607 1.0000 cb 0.0374 -0.0308 -0.0226 0.2258 0.0807 1.0000

Before including all variables in the regression formula, the correlations between them should be examined. It can be seen in Table 3 that all variables correlate with the dependent variable

13

except the cb variable. Probably this is partly due to the erased observations from the dataset, because the correlation of cb with all other variables is low, except for gdpindex. It could also be that the amount of contributions and benefits does not immediately affect the funding ratio. However, the cb variable will not be included in the regression. The variables that remain result in the following formula which will be tested:

𝑐𝑜𝑣𝑒𝑟𝑟𝑎𝑡𝑖𝑜_𝑖,𝑗 = 𝛼 + 𝛽₁𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡𝑟𝑎𝑡𝑒_𝑗+ 𝛽₂𝑟_𝑖,𝑗+ 𝛽₃𝑛100_𝑗+ 𝛽₄𝑔𝑑𝑝𝑖𝑛𝑑𝑒𝑥_𝑖,𝑗+ 𝜀_𝑖,𝑗

The theory behind this formula indicates that the QE program of the ECB will lower the interest rate. A lower interest rate will increase the market value of the liabilities of pension funds. Because pension funds have a negative duration gap, the market value of liabilities is expected to increase more than the market value of assets. Thereby the cover ratio will decrease. Therefore, it is expected that a lower interest rate results in a lower cover ratio:

H0: 𝛽₁=0 H1: 𝛽₁>0

As research method for the regression of the panel data Generalized Least Squares (GLS) with random effects will be used. Because there are no omitted variables that correlate with the dependent variable, it is expected that the error term does not correlate with the variable

interestrate. Therefore, using random effects will lead to unbiased estimates. The GLS with

random effects will be used because the error term does not correlate with the dependent variables and it is also not expected that there is a correlation between the residuals.

14

4. Results

In this section the results will be given and discussed. The hypothesis that a lower interest rate leads to a lower cover ratio will be tested. For the variables a t-test will be used to conclude whether they are significant. Hereby, a type I error of 5% will be used. After performing the regressions one for one, adding a control variable each time, the results change significant ly. First, the changes in the values of the regressions will be discussed. Thereafter, the hypothesis will be tested. Finally, the outcome will be linked to the general literature.

Table 3. Regression output

Model (1) Model (2) Model (3) Model (4)

coverratio coverratio Coverratio coverratio

interestrate -2.66*** -2.52*** -2.10*** -1.65* (0.77) (0.80) (0.80) (0.96) r 0.029*** 0.028*** 0.030*** (0.0084) (0.0083) (0.0070) n100 0.18** 0.21** (0.085) (0.092) gdpindex 0.0012 (0.0012) _cons 1.13*** 1.08*** 1.06*** 0.93*** (0.032) (0.027) (0.027) (0.14) N 62 61 61 61 r2_o 0.075 0.42 0.45 0.44

Standard errors in parentheses

* p<0.10, ** p<0.05, *** p<0.01

Model (1) includes the independent variable interest rate. In model (2) the control variable risk-taking behaviour is added. Model (3) also includes the control variable for equity prices, n100. Model (4) contains also the GDP control variable.

15

Table 3 shows the results of the regressions. For each model there has been added one control variable to see how the results change. When running a regression only with the main explanatory variable, the outcome of 𝛽₁ is significant with p<0.01. However, the effect is exactly the opposite as the expectation. 𝛽1 is namely negative instead of positive, which would

result in a higher cover ratio when the interest rate decreases. This would indicate that the QE program has a positive effect on pension funds. Before drawing any conclusions, the control variables and the other models should be taken into account.

The control variable r for risk taking behaviour of pension funds is significant in all models. With a value between 0.028-0.030 and a significance of p<0.01 it can be concluded that a portfolio with more risky investments leads to a higher cover ratio for pension funds. Also, it can be concluded from the table that the value of 𝛽₁ increases after adding the risk-taking behaviour variable. Furthermore, the R2 _{increases from 0.0748 to 0.4230 in the second} model. The variable n100 is also significant with p<0.05. Adding this variable also increases the explanatory power to 0.44. Including GDP decreases the explanatory power of the model and the variable itself is not significant. Moreover, the interestrate loses significance when

gdpindex is added. This might be due to the high correlation between interestrate and gdpindex.

N100 and GDP were both chosen as variables to control for economic and financial shocks, such as the financial crisis of 2007. GDP might not have the expected effect because it differs to much between countries, while the pension funds depend more on global economic trends. All in all, the models seem to explain the influences on the cover ratio because the variables are significant. However, there is not enough evidence to accept the hypothesis that a lower interest rates leads to a lower cover ratio. In contrast, evidence in exactly the opposite direction has been found. A lower interest rate seems to result in a higher cover ratio. Future research should verify whether this is the case. Focussing on the theory, the expected effect has not been found. This might be attributed to theoretical implications, data implications or method implications.

The theory shows how a lower interest rate leads to a lower cover rate. The interest rate is used for discounting liabilities and assets. A lower interest rate leads to an increase in the market value of liabilities and assets. Because most pension funds have a negative duration gap, their liabilities would increase more than their assets. The opposite effect found in this analysis might mean that pension funds do not have a negative duration gap at all. Although this might sound unrealistic, it could be that pension funds hedge their duration risk. Hedging the risk would mean that the market value of liabilities does not increase more. It could also be

16

that the buying of long-term bonds by pension funds, argued by Domansky, Shin and Sushko (2017), resulted in a non-negative duration gap. Although they expected that this buying of governments bond would lead to the adverse effect, it could be that pension funds had a positive duration gap. The cover ratio would then increase when the interest rate decreases. Another theoretical explanation is that the cover ratio would increase because the market value of the assets can increase by a lower interest rate. It can be useful to investigate whether pension funds use hedging technics to cover their duration mismatch. This could explain the negative effect from the interest rate on the cover ratio.

Besides possible theoretical explanations for the unexpected effect of the interest rate on the cover ratio, the result could also be unreliable because of the dataset. The amount of observations might not be large enough. Although this cannot directly be related to the negative result, the insignificance of gdpindex, and the interestrate losing significance after adding more control variables, might be caused by having not enough observations. Additionally, the dataset from EIOPA consists of countries which independently gathered their data. This could have led to accounting differences. Therefore, not all observations might be reliable. Finally, the data consists of yearly averages per country, which might not contain enough variation to get reliable results.

Methodological issues concerning the results point mostly towards the time span. The negative effect interest rate has on cover ratio, can be caused by the recovery of the financial crisis. When gdpindex and n100 cannot fully control for this recovery, an increase in the cover ratio could be attributed to an increase in the interest rate. In the first year of the dataset, 2010, the effects of the financial crisis are expected to have maintained for quite some years. Despite controlling for these effects, probably not all effects were captured by the control variables. Therefore, the research should be conducted on a time span starting in a later year. Due to data availability, this will only be possible on a later date.

17

5. Conclusion

The current literature about the effect of low interest rates on pension funds provides us with some useful insights but did not cover the whole field. Boubaker et al. (2017) show that the monetary policy conducted by the Fed after the financial crisis in 2007 made pension funds rebalance their portfolio, investing more in risky assets because of the low interest rate. This indicates that low interest rate might harm pension funds when they do not partake in more risk taking. Domanski, Shin and Sushko (2017) argue that European funds did not take more risk after the ECB announced their QE program, but instead started buying more long-term government bonds. Because of a difference in accounting rules, the European pension funds can not undertake more risk and would therefore be more vulnerable to a decrease in the interest rate. This research aimed to investigate whether a lower interest rate leads to a lower cover ratio of pension funds.

The analysis provided mixed results. The results did not support the hypothesis which states that a lower interest rate leads to a lower cover ratio. To the contrary, the results indicated an opposite effect of the interest on the cover ratio. The regressions indicated that a lower interest rate would lead to a higher cover ratio. This means that the QE program would not harm pension funds. The unexpected effect can be caused by the dataset. The dataset may not contain enough observations, and the observations were probably subject to accounting differences. Furthermore, the financial crisis in 2007 could have led to an opposite effect as the interest rate. The economy was still recovering, which could have led to higher cover ratios from pension funds. Although the regression contained variables to control for these effects, they might not have captured everything. Because the results were significant, the unexpected effects are probably due to the latter explanation. However, the interest rate substantially lost significance by adding more control variables while the explanatory power of the model increased. Finally, pension funds have reacted on the QE program by buying long-term bonds. Thereby decreasing their negative duration gap. If pension funds did not have a negative duration gap, a lower interest rate would not have decreased the cover ratio.

All in all, this research contributes to the current literature because it focused on European pension funds which are long term savers without the ability to undertake more risk. Although this paper does not show that the QE program of the ECB harms pension funds, it does indicate that there is a relationship between the interest rate and the financial position of pension funds. The main limitation of this research was the availability of data. Because there is no data available on pension funds in Europe before the financial crisis, it was not possible

18

to exclude it. Future research should use a data set starting on a later date, thereby avoiding the influence of the financial crisis. Furthermore, authorities such as EIOPA and the Dutch central bank (DNB) started to collect more data on pension funds a few years ago. The DNB also collects quarterly data on individual pension funds. Using such data contains more variation, thus analysis of this data can give more reliable results. Besides that, research should focus on the duration gap of pension funds. Showing that pension funds have a negative duration gap underlines the theoretical assumptions of this paper.

19

Reference list

Andonov, A., Bauer, R., & Cremers, M. (2017). Pension Fund Asset Allocation and Liability Discount Rates (February 21, 2017). Available at SSRN:

https://ssrn.com/abstract=2070054or http://dx.doi.org/10.2139/ssrn.2070054

Antolin, P.S. & Yermo, J. (2011). The Economic Impact of Protracted Low Interest Rates on Pension Funds and Insurance Companies. OECD Journal: Financial Market Trends,

2011(1), 237- 256.

Bikker, J.A., Knaap, T., Romp, W.E. (2014). Measuring and explaining implicit risk sharing in defined benefit pension funds. Applied Economics, 46(17), 1996-2009.

Boon, L.N., Brière, M., & Rigot, S. (2018). Regulation and pension fund risk-taking. Journal

of International Money and Finance, 84, 23-41.

Boubaker, S., Gounopoulos, D., Nguyen, D.K., & Paltalidis, N. (2017). Assessing the effects of unconventional monetary policy and low interest rates on pension fund risk

incentives. Journal of Banking and Finance, 77, 35-52

Brucciol, A. & Beetsma, R.M.W.J. (2010). Inter- and Intra-generational Consequences of Pension Buffer Policy under Demographic, Financial, and Economic Shocks. CESifo

Economic Studies, 56(3), 366-403.

Claeys G, Leandro A, Mandra A (2015) European Central Bank quantitative easing: The detailed manual. Bruegel Policy Contribution 2015/02, Bruegel, Brussels

Di Maggio, M., Kacperczyk, M. (2017). The unintended consequences of the zero lower bound policy. Journal of Financial Economics, 123(1), 59-80.

Domanski, D., Shin, H., & Sushko, V. (2017) The Hunt for Duration: Not Waving but Drowning? IMF Economic Review, 65(1), 113-153.

Gali, J. (2014). Monetary Policy and Rational Asset Price Bubbles. American Economic

Review, 104(3), 721-752.

Gambacorta, L., Hofmann, B., Peersman, G. (2014). The Effectiveness of Unconventional Monetary Policy at the Zero Lower Bound: A Cross-Country Analysis. Journal of

Money, Credit and Banking, 46(4), 615-642.

Joyce, M., Miles, D., Scott, A., & Vayanos, D. (2012). Quantitative Easing and

Unconventional Monetary Policy – an Introduction*. Economic Journal, 122(564), F271-F288.

Weale, M. & Wieladek, T. (2016). What are the macroeconomic effects of asset purchases?