The impact of defined benefit pension liabilities and costs on the market value of AEX companies

(1)

The impact of defined benefit pension liabilities and costs on

the market value of AEX companies

26 January 2016

MSc in Actuarial Science and Mathematical Finance Thesis Author

L.J.H. Hagens BSc 5800897

Supervisor

Prof. dr. R. J.A Laeven Second reader

(2)

Abstract

AEX 25 companies need to value and report defined benefit pension plans in the financial statements and the accompanying footnotes in accordance with IAS19 under IFRS. In 2011 the IASB published their amendment to IAS19 to address the concerns of analysts and investors relating to transparency in accounting and reporting of defined benefit plans. This amendment was effective for fiscal years beginning on or after 1 January 2013. Restatement of prior periods’ financial statements and financial information was required for comparative purposes. This thesis has looked whether analysts and investors look at the net pension cost, core pension cost and pension liabilities to value a company. There are two underlying models that can be considered: the opaque model and the transparent model. The opaque model assumes that the market only takes into account the core pension cost, where the transparent model only takes into account the net pension assets. This thesis has also looked whether fair value pension liabilities, which are based on government bond yields instead of corporate bond yields, better explain the market value of companies. The same OLS model was used as used by Richardson et al. (2014), which was based on the OLS model as developed by Coronado et al. (2008). Richardson et al. (2014) have done a study on FTSE 100 companies in the period 2006 – 2012, which was before the amendment in IAS19. They showed that the market was transparent, but also that the size of the pension liabilities impacts the market value of the company. They also found that it was more likely that analysts and investors were using the fair value of pension liabilities when valuing a company. In this study on AEX 25 companies in the period 2012 – 2014 it was found that all three pension variables were not to be significant in both the

original dataset and the fair value dataset. This is in contrary with the findings of Richardson et al. (2014). Based on these results I think that although IAS19 has created more transparency and better understanding of the risks of the pension liabilities, it is my opinion that analysts and investors are using other variables to value a company.

(3)

Table of content

1 Introduction 1

2 Literature review 4

2.1 Pension reporting under IFRS 4

2.2 Background last amendment to IAS19 6

2.3 Past articles on the effect of pension information on market value 8

3 Data 11

3.1 Original dataset 11

3.2 Dataset with fair value pension liabilities 14

4 Model 18

4.1 The OLS model 18

4.2 The OLS model as adopted for this study 18

5 Results 21

5.1 Findings 21

5.2 Comparison with findings of Richardson et al. (2014) 24

6 Conclusion and discussion 25

References 28

Appendix A List of AEX 25 companies with Euronext ticker name 29

(4)

1 Introduction

Pension liabilities and net pension liabilities are volatile components of the balance sheets of companies. There are two different types of pension plans; the defined contribution plan and the defined benefit plan or hybrid forms of these types. The defined contribution plan is based on the principle that the employer (and the employee) pays an agreed amount or percentage of the pensionable salary in a pension pot. At the retirement age this pot of accumulated contributions plus any investment returns earned will be used to buy a pension for the employee. Any long-term risks are borne by the employee. The defined benefit plan is based on the principle that the pension at retirement age depends on the years of service and future pensionable salary. These future promises to the employees are the responsibility of the company.

Companies need to value and report the defined benefit liabilities under the financial statements in their annual report using either the accounting standard International Financial Reporting Standards (IFRS), specified in International Accounting Standard 19 (IAS19), or accounting standard United States Generally Accepted Accounting Principles (US GAAP), specified in Accounting Standards Codification 715 (ASC 715). To estimate the present value of the future defined benefit liabilities to current and past employees, actuarial assumptions need to be made, for example on the discount rate, salary increase rate, pension increase rate, mortality and withdrawal, making these volatile and risky liabilities for the company. The assets of pension plans can also be unstable, as they depend on stock market movements. This has caused companies to look for de-risking solutions ranging from using financial instruments and insurance buy-ins on the asset side and the closure and even transfer of the defined benefit plans from final salary based plans to average salary plans, hybrid defined benefit – defined contribution plans or to defined contribution plans to free the company of any risks. In the Netherlands this movement has already taken place and is still going on as also seen by Ponds and Van Riel (2007).

(5)

IAS19 and ASC715 sets out the rules for recognizing employee benefits, such as defined benefit pension plans, in the financial statements and the accompanying

footnotes. Analysts and investors have raised their concerns in the past that the accounting and reporting of the pension liabilities and costs did not give enough transparency around the risks of pension plans and effect on the balance sheet and net income statement. Coronado et al. (2008) pointed out the lack of transparency of ASC715 in the year 1993 - 2005, which caused investors to only focus on the pension cost and not on the net pension liabilities and thereby incorrectly value companies that have defined benefit pension plans. Richardson et al. (2014) did the same study on FTSE 100 companies, who were following the IAS19 accounting rules, for the period 2006 – 2012. Their findings were in contrast to the study done by Coronado et al. (2008), that the market only takes into account the net pension liabilities instead of the pension cost, suggesting that IAS19 accounting rules have a greater market credibility than the ASC715 at that time. They also found that the size of the pension liabilities impacted the market value. Richardson et al. (2014) pointed out that companies can choose their discount rate based on corporate bond yields with similar durations as the pension liabilities, making it the largest systematic bias in pension liability estimates. They therefore also studied the impact of using fair value pension liabilities, which have discount rates set in line with government bonds. These liabilities are higher because government bond yields are usually lower than corporate bond yields. They found in their study that fair value liabilities and net assets seem to estimate market values better, suggesting that analysts and investors apply a risk premium on top on the reported pension liabilities to allow for systematic risk.

In 2011 an amendment was issued to IAS19 to induce more transparency in the pension accounting and reporting that took effect as of 1 January 2013. In the pension footnotes it is required to show comparative years and therefore the amendment is also shown for the fiscal year 2012. The amendment includes the change to immediate recognition of actuarial gains and losses in the estimation of the pension liabilities and assets and by estimating the expected return on assets to be equal to the discount rate. With this thesis I want to see if the amendment in IAS19 has helped analysts and

(6)

investors to better value companies with defined benefit pension plans. The OLS model from Richardson et al. (2014) will be used. I will look at the current (2015) AEX 25 companies for the fiscal years 2012 – 2014, who all report under IAS19. Previous studies have looked at this topic, and drawn interesting conclusions, but these were all before the amendment. However this study looks into this after the amendment to IAS19 to see how effective these amendment has been.

The thesis is constructed as follows. Chapter 1 explains the literature review I have done on the background of IAS19 and the impact of the amendment in 2011 on pension accounting and reporting. I will also study previous research done in the past and their findings. Chapter 2 outlines the data I have used including descriptive statistics. In Chapter 3 explains the OLS model I have used. Chapter 4 sets out my findings. Finally, chapter 5 will be a summary of all the above and will include some points for

(7)

2 Literature review

The chapter will explain the background on IAS19, the amendment that took place in 2011 and the findings of previous researches done on the impact of pension accounting and reporting on market value of companies.

2.1 Pension reporting under IFRS

IAS19 Employee Benefits (last amendment 2011) is an accounting rule concerning employee benefits under the International Financial Reporting Standards (IFRS) rules set by the International Accounting Standards Board, split out in short-term benefits, post-employment benefits such as defined benefits plans, other long-term benefits and termination benefits. The standard establishes the principle that the cost of providing employee benefits should be recognised in the period in which the benefit is earned by the employee, rather than when it is paid or payable. It also outlines how each category of employee benefits is measured, providing detailed guidance in particular about post-employment benefits.

IAS19 splits out long-term employee benefit plans as follows:

 Post-employment plans o Defined benefit plans

 Pension plans

 Post-retirement medical plans o Defined contributions plans

 Other long-term benefit plans, which include jubilee or other long-service benefits, long-term disability benefits and, if they are not fully payable within 12 months after the end of the period, bonuses and deferred compensation.

 Termination benefits, which are benefits offered by a company for terminating the contract or to encourage voluntary terminations due to restructuring and

(8)

should not be mistaken for benefits already specified in the employee contract, either voluntary or involuntary.

Companies are required to report the following in the pension notes:

- More insight on the risks of the largest defined benefit plans and regulatory environment in which the plans operate

- Reconciliation from beginning to end of year of the defined benefit liabilities - Reconciliation from beginning to end of year of the defined benefit assets

- Reconciliation from beginning to end of year of the defined net benefit liabilities - Defined benefit cost/pension cost split out by separate components

- Explanation of any large movements that have occurred - Key weighted average actuarial assumptions used

- Sensitivities on the defined benefit liabilities in the scenario where each key actuarial assumption goes either up or down

- Split of the defined benefit assets by assets category

For all the above except the sensitivities, the figures for the previous year should also be reported as comparative.

Some companies also report some additional information, for example

 Report the figures for an additional year before

 Weighted average duration of the plan

 Split of each reconciliation, pension cost and key weighted average for pension plans and post-retirement medical plans

 Split of the total defined benefit liability for actives (current employees), deferreds (past employees) and pensioners.

Each company can choose their assumptions from a range that is accepted by the auditor depending on the type of plan, duration and market practice.

(9)

Table 1 shows how each component of the defined benefit cost is recognised under IAS19 as stated by IASB (2011).

Table 1 – Recognition of the defined benefit cost items

Defined benefit cost item Recognition

Service cost attributable to the current and past periods, curtailment effects and settlement effects

To be recognised in the income statement as operating costs

Net interest on the net defined benefit liability or asset, determined using the discount rate at the beginning of the period

To be recognised in the income statement as financing costs

Remeasurement effects of the net defined benefit liability or asset, comprising of changes in actuarial assumptions, experiences different than assumed including gain/loss on return on defined benefit assets higher/lower than the discount rate and changes in the effect of the asset ceiling, when applicable.

To be recognised in Other Comprehensive Income as remeasurements

2.2 Background last amendment in IAS19

In 2011 IASB published their amendments to IAS19 to be effective for fiscal years beginning on or after 1 January 2013; retrospective application (restatement of prior periods’ financial statements and financial information presented for comparative purposes) is required. The main driver for these amendments was to address the concerns of users of financial statements about employers’ transparency in accounting and reporting on their defined benefit plans under current accounting rules, the risks associated with their plans and the impact of the plans on the company financial position and income.

(10)

The three most important changes that increased the transparency and impacted the income statement and balance sheet are set out below. For each change the impact is explained.

Before Currently

Actuarial gains and losses could be recognized via a corridor approach, meaning only the gains and losses above a certain amount “corridor” needed to be recognized on a delayed basis. The corridor was defined as 10% of the greater of the DBO or the fair value of plan assets. The gains and losses above the corridor was then divided by the participants’ future service periods to be amortised in the future years

Recognising all changes in net defined benefit liability immediately, of which the pension cost is recognized in the income statement and all actuarial gains and losses to be recognized as

remeasurements in other comprehensive income.

The impact of this change caused companies to have a higher pension liability or lower pension asset in their balance sheet. Also, the net income has become less volatile, but greater volatility in the balance sheet asset or liability. Continuing with the following change:

The expected return on assets

(component in the pension cost) was based on the expected return of each asset allocation, meaning these were usually higher than the discount rate.

The expected return on assets is based on the discount rate at the beginning of the year and has been renamed as interest income on assets.

(11)

The impact of this change cause companies to have a higher pension cost, so a lower net income. Lastly, the final change:

Sensitivity analysis on the defined benefit liabilities for changes in key actuarial assumptions was not required

Sensitivity analysis on the defined benefit liabilities for changes in key actuarial assumptions is required

This change should help analyst and investors understand how a change in key actuarial assumptions can impact the defined benefit liability.

2.3 Past articles on the effect of pension information on market

value

Coronado et al. (2003) and the extended follow-up study, Coronado et al. (2008), studied the effect of pension accounting on market values of S&P 500 companies between 1993 - 2005. Their OLS model was based on the simple residual income model as used by Feltham and Ohlson (1995), where firm equity value is expressed as the sum of the value coming from the firm’s nonfinancial operating activities plus the value of unrelated financial activities. They further developed the model by dividing both the balance sheet and the income statement in the core business operations and the results from pension plan asset/liability management. However, their study was to show the impact of the limited transparency of US GAAP reporting standards in relation to the pension liabilities, pension assets and pension costs. They assumed the market to be either transparent or opaque by using the following model:

MarketCapital = β₁ + β₂ CoreBookValue + β₃ CoreEarnings (1) + β4 NetPensionAssets + β5 CorePensionCost + ε

where

(12)

 CoreBookValue being the book value excluding the book value of the net pension assets

 CoreEarnings being the earnings excluding the core pension costs

 NetPensionAssets being the pension assets minus the pension liabilities

 CorePensionCost being the total annual pension cost minus the service cost. Service cost is the present value of the pension benefits that employees have earned over the year and they therefore consider this to be part of the core labor and wages expense of the company.

 All variables normalized by total company assets to remove time trends and the associated heteroskedasticity

 Year dummies and broad industry dummies that interacted with year dummies to allow for industry-related differences in valuation and growth prospects

 The model also included the variables growth forecast and firm size

 There dataset also included companies with no defined benefit pension plans. Therefore a dummy was included to identify whether a company has a defined pension benefit pension plan or not.

 Using the method proposed by Thomson (2006) they estimated

heteroskedasticity-consistent standard errors by grouping by year and company.

Between 1993 and 2005 US GAAP required companies to disclose in the footnotes the pension liabilities and assets together with the total pension expense split by the cost of the new benefits offset by the expected long-term returns of pension assets. Although it should be noted that companies only needed to reflect the total pension expense in the net income statement. Coronado et al. (2008) made the hypothesis that the turbulence in net pension assets in the period 2000 – 2002 due to low discount rates and low asset returns that drew public attention to defined benefit plans, have made analysts and investors more alert to the value of the net pension assets.

The market can be considered to be transparent when analysts and investors focus on the net pension assets instead of the core pension cost to value the company.

(13)

This would mean that β3 would be non-zero and β4 would be zero. If the situation ws reversed, the market can be considered opaque. Coronado et al. (2008) concluded that the market was opaque in the period 1993 – 2001 and still was in the period 2002 - 2005.

After 2008 reforms have been made to the accounting and presentation of pensions and pension-related information in the financial statements under both US GAAP and IFRS. Richardson et al. (2014) based their study on FTSE 100 companies under

accounting method IFRS in the period 2006 - 2012. They used the exact same model as Coronado et al. (2008). When they tested if the market was transparent or opaque, they concluded the opposite of the findings of Coronado et al. (2008). The market seems to be transparent, meaning that analysts and investors take into account the net pension assets instead of the pension cost. They were claiming that this may be attributed to the fact that IFRS pension accounting have a greater market credibility than US GAAP pension accounting. They broadened their study by investigating whether analysts and investors take the size of the pension liabilities into account. It was concluded that they did. They also noted that pension liabilities have different sources of bias and systematic risk. The use of discount rates is the largest systematic risk the market can take into account. IAS19 states the use of corporate bonds with at least a rating of AA to set the discount rate based on bond yields with similar durations. However, Richardson et al. (2014) thinks that fair value pension liabilities should be based on the government bond yields. They used the discount rate sensitivity information in the pension notes to derive the duration. They found that these pension liabilities are higher because government bond yields are usually lower than corporate bond yields. They applied the same model on the fair value liabilities and net assets and concluded that the model had a higher explanatory power.

(14)

3 Data

This chapter will explain the two datasets that were used in this study. The first section will focus on the original dataset, followed by the section that focus on the dataset including fair value estimates.

3.1 Original dataset

The dataset in this study consists of 61 data points. The financial and pension information of the current (2015) AEX25 companies for the period 2012 – 2014 (3 years) was collected. The list of AEX25 companies can be found in Appendix A. Table 2 shows the list of companies included in the dataset. Altice, OCI and Unibail-Rodamco were not included in the dataset due to not reporting any pension liabilities and costs for the whole period. ASML has no obligation to pay off any deficits to the pension fund or claim any potential surpluses, so these are accounted for in a similar way as a defined contribution plan in the financial statements and was therefore also excluded from the dataset. For NN Group only the 2014 figures were available as they entered the AEX (Euronext) in July 2014.

An overview of each variable and the source of the information can be found in Appendix B. The pension liabilities, pension assets and core pension cost (total pension cost excluding service cost) were taken by hand from the pension notes linked to the financial statements in the annual reports. Market capital, book value and earnings were taken from Bloomberg. Core.BookValue is the book value minus the net pension assets. Core.Earnings is the earnings minus the core pension cost. Table 3 shows the

descriptive statistics of the original dataset. In cases where the annual report or Bloomberg reported the financial information in a currency other than Euro, the information was converted using the ask price exchange rate from www.oanda.com

(15)

Table 2 – List of companies included in the dataset Company 2012 2013 2014 Aalberts Industries X X X Aegon X X X Ahold Koninklijke X X X Akzo Nobel X X X Altice ArcelorMittal X X X ASML Holding

Boskalis Westminster Koninklijke X X X

Delta Lloyd X X X DSM Koninklijke X X X Gemalto X X X Heineken X X X ING Groep X X X KPN Koninklijke X X X NN Group X OCI Philips Koninklijke X X X Randstad X X X RELX X X X

Royal Dutch Shell A X X X

TNT Express X X X

Unibail-Rodamco

Unilever Certificate X X X

Vopak Koninklijke X X X

Wolters Kluwer X X X

Only five companies reported pension liabilities lower than pension assets and therefore have positive net pension assets. As seen in table 3 it can be noted that the range of each variable is very wide, where the maximum can be more than 10 times than the minimum

Only pension liabilities and costs were taken into account, meaning that the liabilities and costs for the post-retirement medical plans and other long-term benefit

(16)

plans were excluded. The reason for this is that the post-retirement medical plans and the other long-term benefit plans are usually unfunded, because they don’t have to be funded, and are usually smaller than pension plans, therefore do not reflect a large risk for the company, where the pension liabilities are. In some cases there was no split possible and therefore all defined benefit plans were taken into account. As mentioned the amount of liabilities and costs attributed to these plans is relatively small compared to the pension plans, so it should not affect the dataset significantly.

Table 3 – Descriptive statistics of the original dataset (in millions EUR)

========================================================================

Statistic N Mean St. Dev. Min Max

--- Pension.Assets 61 8,627.50 14,395.10 9 71,083.00 Pension.Liabilities 61 -10,068.50 16,288.80 -83,446.00 -96.80 Net.Pension.Assets 61 -1,441.00 2,684.20 -12,363.00 607 Core.Pension.Cost 61 -35.30 201.10 -874 640 Core.Bookvalue 61 17,507.30 32,224.40 1,035.00 160,631.50 Core.Earnings 61 1,599.00 3,789.60 -1,409.90 21,610.30 Total.Assets 61 106,913.20 243,660.90 1,955.50 1,166,191.00 Market.Capital 61 23,619.90 37,985.60 1,717.60 185,844.30 Bookvalue 61 16,066.30 30,110.10 968.9 148,268.50 Earnings 61 1,563.70 3,776.50 -1,500.40 21,219.30 --- ---

Additional notes on the pension notes

 When multiple pension plans were reported, aggregated figures were used.

 Each annual report has a different lay-out and uses different synonyms.

 It has occurred that discount rate sensitivity information in 2012 was only

available in accordance with IAS19 before the amendment or not available at all. As the pension liabilities in 2012 do not differ materially before and after the

(17)

amendment, I have used the information available. Where no information was available, a similar duration as in 2013 was assumed for pragmatic reasons.

3.2 Dataset with fair value pension liabilities

The discount rate used by companies to value their pension liabilities is chosen from a range, often advised by their actuary and audited by their auditor. This can vary

depending on plan type, duration of the pension liabilities and market practice, provided that the underlying method to set the discount rate remains consistent (unless if there is a good reason for it to change). This causes a different set of discount rates when comparing across companies and over time. IAS19 states that the uses of AA corporate bond yields are seen as high quality bonds to set the discount rate. However, current market rates on government bonds are usually lower than corporate bond yields, therefore the fair value pension liabilities will be higher.

To estimate the fair value of the pension liabilities, the implied duration is needed. Although the pension notes mention the weighted average duration of the defined benefit plans or pension liabilities only, this information was not used, because the reported weighted average duration is based on the timing and size of future

expected cash flows. The implied duration is based on the 1% and -1% effect of change in discount rate. IAS19 states that discount rate sensitivities should be provided and can therefore be found in the pension notes.

𝐼𝑚𝑝𝑙𝑖𝑒𝑑 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛 =ln (𝑝𝑒𝑛𝑠𝑖𝑜𝑛 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠𝑢𝑝 𝑋%⁄𝑝𝑒𝑛𝑠𝑖𝑜𝑛 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠𝑑𝑜𝑤𝑛 𝑋%)

ln ((1+𝑑𝑖𝑠𝑐𝑜𝑢𝑛𝑡 𝑟𝑎𝑡𝑒𝑑𝑜𝑤𝑛 𝑋%) (1+𝑑𝑖𝑠𝑐𝑜𝑢𝑛𝑡 𝑟𝑎𝑡𝑒⁄ 𝑢𝑝 𝑋%)) (2)

For TNT Express the fair value liabilities could not be estimated due to missing

discount rate sensitivity information in the annual reports each year. Therefore there are only 58 (61 – 3) data points. Table 4 outlines the average implied duration per year with the related average discount rate. AEX 25 companies have most of their pension

liabilities in the Eurozone. When these were not in the Eurozone, the pension liabilities were often in either in the UK or US or in both countries. Non-Eurozone discount rates

(18)

are usually higher, which causes the weighted average discount rates to be slightly higher than for companies with pension liabilities only in the Eurozone.

Table 4 – Average implied duration of the pension plans per year and the accompanying average discount rate used

Year Average implied duration Average discount rate

2012 14.3 3.87%

2013 14.1 4.04%

2014 14.8 2.79%

For comparison I have looked at the Markit iBoxx Eurozone AA corporate bond yields with a duration of more than 10 years. These were taken from Bloomberg. These bond yields increased between 31 December 2012 and 31 December 2013 with 48 basis points. Between 31 December 2013 and 31 December 2014 these bond yields decreased with 168 basis points. This movement can also be noticed in the movement of the average discount rate in combination with the average implied duration over the years.

Eurozone government bond yields with similar durations were used to estimate the fair value pension liabilities.

𝐹𝑎𝑖𝑟 𝑣𝑎𝑙𝑢𝑒 𝑝𝑒𝑛𝑠𝑖𝑜𝑛 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑒𝑠 = 𝑝𝑒𝑛𝑠𝑖𝑜𝑛 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 ∗ (_{1+𝑑𝑖𝑠𝑐𝑜𝑢𝑛𝑡 𝑟𝑎𝑡𝑒}1+𝑑𝑖𝑠𝑐𝑜𝑢𝑛𝑡 𝑟𝑎𝑡𝑒

𝑔𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑏𝑜𝑛𝑑 𝑦𝑖𝑒𝑙𝑑𝑠)

𝑖𝑚𝑝𝑙𝑖𝑒𝑑 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛

(3)

The Markit iBoxx Eurozone government bond yields were taken from Bloomberg. Table 5 shows the Markit iBoxx Eurozone government bond yields as at 31 December for the years 2012 – 2014 per duration.

(19)

Table 5 – Markit iBoxx Eurozone bond yields per duration and date

Duration 31 December 2012 31 December 2013 31 December 2014

1-3 years 1.01% 0.78% 0.18%

3-5 years 1.27% 1.34% 0.32%

5-7 years 1.98% 2.05% 0.58%

7-10 years 2.40% 2.58% 0.91%

10+ years 3.19% 3.50% 1.80%

Where it was mentioned in the pension notes that the pension liabilities includes pension liabilities from countries other than the Eurozone, I have also used the

Eurozone government bond yields for pragmatic reasons. As AEX 25 companies have most of their pension liabilities in the Eurozone, this pragmatic solution should only slightly impact the fair value liabilities, causing them to be lower than they actually would be if the government bonds of the specific non-Eurozone countries were used. Table 6 shows the descriptive statistics of the dataset with the fair value estimates.

Table 6 – Descriptive statistics of the dataset with the fair value estimates (in millions EUR)

========================================================================

Statistic N Mean St. Dev. Min Max

--- Pension.Assets 58 9,043.40 14,647.40 9 71,083.00 FVNet.Pension.Liabilities 58 -2,860.00 5,030.30 -29,344.10 -12.8 FV.Pension.Liabilities 58 -11,903.50 18,851.80 -100,427.10 -102.6 Core.Pension.Cost 58 -37.1 206.1 -874 640 Core.Bookvalue 58 18,281.20 32,873.60 1,035.00 160,631.50 Core.Earnings 58 1,678.60 3,871.10 -1,409.90 21,610.30 Total.Assets 58 112,219.30 248,823.00 1,955.50 1,166,191.00 Market.Capital 58 24,646.90 38,691.20 1,717.60 185,844.30 Bookvalue.in.millions 58 16,773.10 30,724.60 968.9 148,268.50 Earnings.in.millions 58 1,641.50 3,858.40 -1,500.40 21,219.30 --- ---

(20)

The dataset only excludes TNT Express and therefore the statistics does not differ much from the whole dataset. Table 7 shows the ratio fair value figures versus reported

figures by years. The fair value pension liabilities for all observations were on average 11% higher than the reported pension liabilities. The fair value of the net pension liabilities for all observations were on average 56% higher than the reported net pension liabilities.

Table 7 – Ratio fair value estimates vs reported pension liabilities and net pension assets

Year Ratio fair value vs reported

pension liabilities

Ratio fair value vs reported net pension assets

2012 109% 143%

2013 107% 112%

(21)

4 Model

This chapter will give background on a regular OLS model and then explain the model which was used in this thesis. The modelling has been done in R software. This

software is accessible for anyone online for free. Separate ‘packages’ can be

downloaded depending on the specific use of R. In this thesis the gdata and stargazer packages were used.

4.1 The OLS model

The ordinary least squares (OLS) model looks as follows:

Y = Xβ + ε (4)

where

 Y being the dependent variable vector (y1, … , yn)

 X being the independent variable matrix (

1 x₂₁ ⋯ x_k1

⋮ ⋮ ⋮

1 x_2n ⋯ x_kn). It is assumed that 𝑛 ≥ k and that matrix X has rank 𝑘.

 β being the unknown parameters vector (β₁, … , β_k), with β₁ being the constant term

 ε being the unobserved disturbances vector (ε1, … , εn) and are independent and normally distributed with mean 0 and standard error of 𝜎2

 𝑘 being the number of unknown parameters

 n being the number of observations

4.2 The OLS model as adopted for this study

The OLS model used in this study is similar to the OLS model used by Richardson et al. (2014).

MarketCapital = β1 + β2 CoreBookValue + β3 CoreEarnings (5) + β4 NetPensionAssets + β5 CorePensionCost + β6 PensionLiab + ε

(22)

CoreBookValue and CoreEarnings represent the non-pension variables, and

NetPensionAssets, CorePensionCost and PensionLiab represent the pension variables. The dependent and the independent variables have been normalized by dividing each variable by the total assets to correct for the heteroscedasticity and make the model more stable. Table 6 shows the description of each variable that was used in the model.

Table 8 – Description of each variable of the model

Variable Description

MarketCapital MarketCapital is the market value of the company divided by total company assets. The market capital is calculated by multiplying share price by number of shares outstanding.

CoreBookValue CoreBookValue is the book value of the company excluding the pension liabilities divided by total company assets. The book value is calculated by multiplying book value per share by number of shares outstanding.

CoreEarnings CoreEarnings is the earnings of the company excluding the core pension cost divided by total company assets. The earnings is calculated by multiplying earning per share by number of shares outstanding.

NetPensionAssets NetPensionCost is the net pension assets (pension liabilities minus pension assets) divided by total company assets. CorePensionCost CorePensionCost is the core pension cost (total pension cost

minus service cost) divided by total company assets.

PensionLiab PensionLiab is the pension liabilities divided by total company assets.

Table 9 shows that the net pension assets and core pension costs are disconnected in sign and size, and correlated differently over time. This is in line with the findings of Coronado et al. (2008). It is expected that these are not always correlated positively with

(23)

the knowledge that core pension cost can move differently than net pension assets as the interest cost on assets depends on the assets the year before and that a one-off

curtailment, settlement or past service cost/credit can take place during the year.

Table 9 – Correlation of net pension assets and core pension cost per year (in millions EUR)

Year Net pension assets Core pension cost Correlation

2012 -205,627 -405 0.5886

2013 -225,553 20 0.9097

2014 -183,000 -1,766 -0.3129

The method of Thompson (2011) to estimate the heteroskedasticity-consistent standard errors by clustering by company and year was not used in this study, as this method only works well when there are at least 25 companies and time periods.

(24)

5 Results

This chapter will first describe the findings of the model. Thereafter, these finding will be compared with the findings of Richardson et al. (2014).

5.1 Findings

Table 10 shows the regression of market value of companies on book value, earnings and three measures of pension value. Standard errors are shown below the coefficient estimates. The signs *, ** and *** indicate the statistical significance level at 10%, 5% and 1%, respectively. The following can be observed:

 Column (1) only includes the effect of the net pension assets. The coefficient is negative and not significant at a significance level of 10%. I would expect the market value to increase for each unit of increase in net pension assets.

 Column (2) only includes the effect of the core pension. The coefficient is positive and not significant at a significance level of 10%. I would expect the market value to increase for each unit of decrease in core pension cost. It can be observed that the prediction power of the model increases slightly when looking at the adjusted 𝑅2_.

 Column (3) only includes the effect of the core pension liabilities. The

coefficient is negative and close to zero. Also the variable is not significant at a significance level of 10%. I would expect the market value to increase for each unit of decrease in pension liabilities. It can be observed that the prediction power of the model is the lowest compared to the previous two models when looking at the adjusted 𝑅2_{. This means that nothing can be said about the impact} of the size of the pension liabilities on the market value.

 It can be observed in column (4) that including both net pension assets and core pension cost as a variable increases the prediction power of the model slightly when looking at the adjusted R2_{. However, this can be attributed to the fact that} adding more variables has led to overfitting the model. Again these coefficients

(25)

are not significant at a significance level of 10%. This means that no conclusion can be made whether the market can be perceived as transparent or opaque. It could be suggested that analysts and investors don’t take both variables into account when valuing a company.

 Column (5) shows the regression when the variable pension liabilities is

included. Again the coefficient is not significant at a significance level of 10%. It should be noted that the sign of the coefficient has changed and that the adjusted 𝑅2_{has decreased. It can be concluded that adding the pension liabilities} as a variable does not increases the model fit.

 Time dummies have been excluded as they were not significant when they were included in the models.

 It can be observed in column 6 until 10 that the same findings hold for the models with the fair value estimates. The only difference is that the models with the fair value estimates have a slightly higher explanatory power when looking at the adjusted 𝑅2_.

(26)

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) CoreBookValue 0.695*** _0.766*** _0.758*** _0.671*** _0.660** _0.572** _0.614** _0.603** _0.572** _0.554** (0.251) (0.225) (0.229) (0.248) (0.256) (0.253) (0.246) (0.250) (0.251) (0.256) CoreEarnings 12.773*** _12.960*** _12.578*** _13.169*** _13.244*** _13.307*** _13.768*** _13.330*** _13.641*** _13.713*** (1.309) (1.298) (1.315) (1.320) (1.388) (1.401) (1.395) (1.413) (1.405) (1.425) NetPensionAssets -1.088 -1.542 -1.755 (1.667) (1.674) (2.018) CorePensionCost 15.255 17.078 17.386 15.861 16.230 16.693 (10.973) (11.165) (11.375) (11.079) (11.107) (11.243) PensionLiab -0.055 0.034 (0.149) (0.177) FVNetPensionAssets -0.533 -0.568 -0.956 (0.640) (0.634) (1.094) FVPensionLiab -0.066 0.098 (0.132) (0.224) Constant 0.003 0.004 0.007 -0.016 -0.013 0.012 0.012 0.013 -0.002 0.005 (0.090) (0.087) (0.091) (0.090) (0.092) (0.090) (0.088) (0.092) (0.089) (0.092) Observations 61 61 61 61 61 58 58 58 58 58 R2 _0.739 _0.746 _0.738 _0.750 _0.750 _0.751 _0.757 _0.749 _0.761 _0.762 Adjusted R2 _0.726 _0.733 _0.724 _0.732 _0.727 _0.737 _0.744 _0.735 _0.743 _0.739 Residual Std. Error 0.298 (df = 57) 0.294 (df = 57) 0.298 (df = 57) 0.294 (df = 56) 0.297 (df = 55) 0.298 (df = 54) 0.295 (df = 54) 0.300 (df = 54) 0.295 (df = 53) 0.298 (df = 52) F Statistic 53.928 ***_{(df =} 3; 57) 55.841***_{(df =} 3; 57) 53.558***_{(df =} 3; 57) 41.982***_{(df =} 4; 56) 33.015***_{(df =} 5; 55) 54.362***_{(df =} 3; 54) 56.155***_{(df =} 3; 54) 53.772***_{(df =} 3; 54) 42.163***_{(df =} 4; 53) 33.254***_{(df =} 5; 52)

(27)

5.2 Comparison with findings of Richardson et al. (2014)

Looking at the results compared to the results found by Richardson et al. (2014), it should be noted that their results were based on a period before the changes in IAS19 and on a different set of companies. Also, their dataset was significantly larger. Still there are some grounds where my results can be compared with their results:

 In line with their results the core book value of my research is significant with a similar size as in their findings.

 The same holds for the core earnings, which is also significant and of similar size. It also has a significantly stronger explanatory impact than the book value of the company.

 They concluded that the market is transparent, which cannot be confirmed by my research.

 The size of the pension liabilities does not seem to impact the market value, where they found this effect.

 In line with their findings the models with the fair value estimates have a slightly higher explanatory power when looking at the adjusted R2. It is unclear how this can be interpreted when all the pension variables are not significant.

IAS19 has caused for more transparency and better understanding of the risks of the pension liabilities. Based on the findings of my research it is my opinion that analysts and investors are using other variables to value a company.

(28)

6 Conclusion and discussion

Defined benefit pension plans are volatile and risky liabilities for the company on both the balance sheet and the income statement. In the past previous studies has shown that measures of pension value are used by analysts and investors to value a company. There are two underlying models that can be considered: the opaque model and the transparent model. The opaque model assumes that the market only takes into account the core pension cost, where the transparent model only takes into account the net pension assets. A study done on S&P 500 companies in the period 1993 – 2005 that reported under US GAAP showed that the market was opaque (Coronado et al., 2003 and

Coronado et al., 2008). A more recent study done on FTSE 100 companies in the period 2006 – 2012 that reported under IFRS showed that the market was transparent, but also that the size of the pension liabilities impacts the market value of the company,

suggesting that analysts and investors also take this into account when valuing a

company. It was also found that it was more likely that analysts and investors use a rule of dumb method to estimate the fair value of pension liabilities to account for systematic bias (Richardson et al., 2014). The systematic bias is caused by IFRS allowing

companies to use AA+ corporate bond yields instead of government bonds yields, and these usually have higher yields which results in reported pension liabilities to be lower than fair value pension liabilities.

To address the concerns relating to transparency in accounting and reporting of defined benefit plans, IASB published their amendment to IAS19, which sets out the rules for recognizing employee benefits, such as defined benefit pension plans, in the financial statements and the accompanying footnotes, in 2011 to be effective for fiscal years beginning on or after 1 January 2013. Retrospective application (restatement of prior periods’ financial statements and financial information presented for comparative purposes) was required. Of these amendments three of them in particular increased the transparency and impacted the income statement and balance sheet:

(29)

 All changes in net defined benefit liability are recognized immediately, of which the pension cost is recognized in the income statement and all actuarial gains and losses to be recognized as remeasurements in other comprehensive income.

 Interest income on assets (component in the pension cost) is based on the discount rate at the beginning of the year.

 Sensitivity analysis on the defined benefit liabilities for changes in key actuarial assumptions

This thesis studied whether these changes have caused the market to remain

transparent, whether the size of pension liabilities matters and if fair value estimates of pension liabilities better explain the market value of companies in line with the study done by Richardson et al. (2014). The dataset consists of financial information of AEX 25 companies in the period 2012 – 2014, excluding companies where no sufficient information was available. This resulted in 61 data points. The same OLS model was used as used by Richardson et al. (2014), which was based on the OLS model as developed by Coronado et al. (2008). The dependent variable was the market value, with the book value and earnings representing the non-pension independent variables, and net pension assets, core pension cost and pension liabilities represent the pension independent variables. For testing whether fair value estimates better explain the market value of companies, one company was omitted from the original dataset causing the fair value dataset to consist of 58 data points.

The results showed that the book value and earnings were significant, with earnings having a stronger explanatory impact, in line with the findings of Richardson et al. (2014). However, all three pension variables were found not to be significant in both the original dataset and the fair value dataset. This is in contrary with the findings of Richardson et al. (2014). Even including the core pension cost and the pension liabilities each separately to the model were found to be not significant. Based on these results I think that although IAS19 has created more transparency and better understanding of the risks of the pension liabilities, it is my opinion that analysts and investors are using other variables to value a company. However the dataset used in this thesis is smaller

(30)

and this should be tested on a larger scale of data. It would be interesting to see if Richardson et al. could extend their study for the years after the changes in IAS19 to see if they would come to the same conclusion.

(31)

References

Coronado, Julia Lynn, and Steven Alan Sharpe (2003). "Did pension plan accounting contribute to a stock market bubble?" Brookings Papers on Economic Activity 2003.1, 323-371.

Coronado, J., Mitchell, O. S., Sharpe, S. A., & Blake Nesbitt, S. (2008). “Footnotes aren't enough: The impact of pension accounting on stock values.” Journal of pension economics and finance, 7(03), 257-276.

Feltham, G. and Ohlson, J. (1995) “Valuation and Clean Surplus Accounting for Operating and Financial Activities.” Contemporary Accounting Research 11(2), 689– 732.

International Accounting Standards Board. (2011). International financial reporting standards (IFRS's): Including international accounting standards (IAS's) and

interpretations as at. London: International Accounting Standards Board. IAS19, Employee Benefits.

Ponds, E. H., & Van Riel, B. (2007). “The recent evolution of pension funds in the Netherlands: The trend to hybrid DB-DC plans and beyond.” Available at SSRN 964428.

Richardson, Pete, and Luca Larcher (2014). Economics, Independent. “The influence of DB pensions on the market valuation of the Pension Plan Sponsor.”

Thompson, S. B. (2011). “Simple formulas for standard errors that cluster by both firm and time.” Journal of Financial Economics, 99(1), 1-10.

(32)

Appendix A List of AEX 25 companies with Euronext ticker name

AEX 25 companies (2015) Ticker name in Euronext

Aalberts Industries AALB

Aegon AGN

Ahold Koninklijke AH

Akzo Nobel AKZA

Altice* ATC

ArcelorMittal MT

ASML Holding* ASML

Boskalis Westminster Koninklijke BOKA

Delta Lloyd DL

DSM Koninklijke DSM

Gemalto GTO

Heineken HEIA

ING Groep INGA

KPN Koninklijke KPN

NN Group** NN

OCI* OCI

Philips Koninklijke PHIA

Randstad RAND

RELX REN

Royal Dutch Shell A RDSA

TNT Express*** TNTE

Unibail-Rodamco* ULA

Unilever Certificate UNA

Vopak Koninklijke VPK

(33)

* Altice, OCI and Unibail-Rodamco were not included in the dataset due to not reporting any pension liabilities and costs. ASML has no obligation to pay off any deficits to the pension fund or claim any potential surpluses, so these are accounted for in a similar way as a defined contribution plan in the financial statements.

** For NN Group only the 2014 figures were available as they entered the AEX (Euronext) in July 2014.

*** For TNT Express the fair value liabilities could not be estimated due to missing discount rate sensitivity information in each annual report.