The Purpose, Motivation and Agreement Content of The Interest Rate Derivatives Usage as Detailed in AEGON’s 2006 Annual Report
Purpose Instrument Type Motivation Agreement content
Interest rate swap agreements
Swaps To effectively convert certain fixed-rate assets and liabilities to a floating-rate basis (generally to six months or less LIBOR), in order to more closely match the performance of the assets and liabilities within AEGON’s portfolio.
These agreements involve the payment or receipt of fixed-rate interest amounts in exchange for floating-rate interest amounts over the life of the agreement without the exchange of the underlying principal amounts.
Option To effectively convert certain foreign currency fixed-and floating-rate assets and liabilities to US dollar floating-rate assets and liabilities.
These agreements involve the exchange of the underlying principal amounts.
Interest rate swap agreements
Swaps To effectively convert certain variable-rate assets and liabilities to a fixed-rate basis in order to match the performance of the assets and liabilities within AEGON’s portfolio more closely.
These agreements involve the payment or receipt of variable rate interest amounts in exchange for fixed-rate interest amounts over the life of the agreement without the exchange of the underlying principal amounts.
Forward starting interest rate swap agreements
Forwards To hedge the variability in future cash flows associated with the forecasted purchase of fixed-income assets.
These agreements reduce the impact of future interest rate changes on the forecasted transaction.
As cash flow
hedges
Cross currency swaps Future To convert variable foreign currency cash flows into fixed cash flows in local currencies. The cash flows from these hedging instruments are expected to occur over the next 30-35 years.
These agreements involve the exchange of the underlying principal amounts. Immaterial amounts of hedge ineffectiveness were recorded in the income statement during 2006, 2005 and 2004. The amount of deferred gains or losses to be reclassified from equity into net income during the next twelve months is expected to be immaterial.
As net
To swap the currency exposure of the debt instrument to the appropriate functional currency.
To ensure that total capital will reflect currency movements without distorting debt to shareholders’
equity ratios.
943 1131 1307
10944 11196 14075
34%
41%
46%
51% 54% 54%
0 2000 4000 6000 8000 10000 12000 14000 16000
2001 2002 2003 2004 2005 2006
0%
10%
20%
30%
40%
50%
60%
National value
(Unit: US$ billion) %
Extent Participation rate
FIGURE 1: The Participation Rate and Extent of Interest Rate Derivatives Usage by Life Insurance Companies
TABLE 1
Variables and Definitions
Variable Definition Panel A: Endogenous variables
Interest rate derivative participation Participation decision: 1 for interest rate derivative users, 0 otherwise
Interest rate derivative usage Ratio of the year-end notional volume of interest rate derivatives by total assets
Interest rate risk exposure Ratio of the return on the life insurer’s common stock due to a 1% change in interest rates Panel B: Control variables
Leverage Ratio of the book value of total liabilities to the market value of equity Convertible bonds Dummy variable = 1 if the life insurer use convertible bonds, 0 otherwise
Affiliation Dummy variable = 1 if the life insurer affiliates to a financial group, 0 otherwise Cash flow Ratio of the cash flow per share scaled by total assets
Firm size Natural logarithm of total assets
Floating rate debt Ratio of the floating rate debt to total long-term debt
Interest coverage ratio Ratio of the operation income before depreciation to the interest expense Quick ratio Ratio of quick assets to current liabilities
Underinvestment costs Ratio of the book value of equity capital to the market value of equity capital
Asset-Liability management Dummy variable = 1 if the life insurer uses the balance-sheet to the interest rate risk management, 0 otherwise
TABLE 2
Descriptive Statistics of The Users and Non-users The Interest Rate Derivatives by Life Insurers and Correlation Matrix
Users (N = 133) Non-users (N = 111)
Variable Mean Std. Dev. Min. Max. Mean Std. Dev. Min. Max. Means of the Wilcoxon signed-rank test
Panel A: Descriptive statistics
Interest rate risk exposure 0.2491 0.2957 0.0491 1.9366 0.0000 0.2491 0.2491 0.2491 12.4502 (0.0000)
Leverage 17.8024 15.6271 0.9286 82.1745 9.0571 8.2287 1.2164 48.3368 12.2905 (0.0000)
Convertible bonds 0.7803 0.4156 0.0000 1.0000 0.6857 0.4665 0.0000 1.0000 2.8483 (0.0044)
Affiliation 0.4773 0.5014 0.0000 1.0000 0.0571 0.4972 0.0000 1.0000 5.7151 (0.0000)
Cash flow 0.0486 0.0569 0.0006 0.5407 0.0569 0.5002 0.0015 0.2305 1.4366 (0.0514)
Firm size 10.8156 2.3959 0.0000 13.7947 8.1989 2.2481 0.0000 11.8874 12.0068 (0.0000)
Floating rate debt 2.9934 2.6448 0.0000 9.8504 3.7289 3.6115 0.0000 15.4137 10.6255 (0.0000)
Interest coverage ratio 4.0632 4.6193 0.0000 28.9301 2.9025 2.8901 0.0000 10.8900 9.9374 (0.0000)
Quick ratio 11.9673 12.3630 0.3830 56.7513 14.3755 15.5462 0.0100 78.8000 12.1309 (0.0000)
Underinvestment costs 12.1923 7.9557 0.0101 46.6071 9.7184 12.9953 0.0084 71.3901 11.6521 (0.0000) Asset-Liability management 0.5935 0.4932 0.0000 1.0000 0.6191 0.4880 0.0000 1.0000 0.2250 (0.0823) Panel B: Correlation matrix
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Interest rate derivative participation (1) —
Interest rate derivative usage (2) - 0.4751* —
Interest rate risk exposure (3) - 0.0919* - 0.0721* —
Leverage (4) - 0.3133* - 0.1263* - 0.0248 —
Convertible bonds (5) - 0.0312* - 0.0044 - 0.0808 - 0.1746 —
Affiliation (6) - 0.1449* - 0.1348** - 0.0896 - 0.1281* - 0.0775 — Cash flow (7) - 0.0221* - 0.2686** - 0.0348 - 0.2019** - 0.1151 - 0.0381* —
Firm size (8) - 0.4775* - 0.3014** - 0.0281** - 0.1275 - 0.4278** - 0.3012* - 0.0353* —
Floating rate debt (9) - 0.1285 - 0.0198 - 0.0056** - 0.0352 - 0.1295 - 0.1102 - 0.1045 - 0.2006** —
Interest coverage ratio (10) - 0.1275 - 0.1156 - 0.0102* - 0.1946 - 0.0852 - 0.0266 - 0.0042 - 0.0026* - 0.0556* — Quick ratio (11) - 0.0756 - 0.0471 - 0.0054* - 0.0969 - 0.1556 - 0.0959 - 0.0512 - 0.0499 - 0.0289 - 0.1081* —
Underinvestment costs (12) - 0.1428 - 0.0403 - 0.0116* - 0.0452 - 0.0551 - 0.1719 - 0.0381 - 0.0956 - 0.0297 - 0.0091* - 0.0933 —
Asset-Liability management (13) - 0.0787 - 0.1849** - 0.0875** - 0.0677 - 0.0169 - 0.0802 - 0.1063 - 0.0501** - 0.0902 - 0.0949 - 0.0445* 0.2668** — Note: Panel A separately reports the descriptive statistics for all independent variables between users and non-users, and reports the differences in the means of user and non-user groups, as well as a nonparametric Wilcoxon signed-rank test of the differences between the distributions. Panel B reports the pair wise of the Pearson correlation matrix for all variables. ***, **, and * represent statistical significance at the 0.01, 0.05, and 0.1 levels, respectively.
26
TABLE 3
Relation between Interest Rate Derivative Participation and Risk Exposure
Dependent variable = Interest rate derivative participation Dependent variable = Interest rate risk exposure
Independent variable Excepted sign PROBIT TSLS VIF HTSR OLS TSLS VIF Interest rate risk exposure + - 0.000927**
( 0.0425 )
Interest coverage ratio - - --1.342496*
--( 0.0681 )
Underinvestment costs - --5.709654
--( 0.1454 )
- 6.463847 ( 0.1672)
- 5.463575
( 0.1423 ) 1.232
Asset-Liability management - --2.976375
--( 0.1355 )
Breusch - Pagan chi-squared – Heteroskedasticity -12.2605 -8.7634 11.2794 9.3421
Notes: Within our analysis, no collinear relationship within our analysis (all calculated VIFs are smaller than 10). Results of the heteroskedasticity test by use the Breusch - Pagan chi-squared test (calculated values: 12.2605, 8.7634, 11.2794 and 9.3421) are smaller than χ2 value (12.59159), mean that we cannot reject the hypothesis of homoskedasticity on this evidence. In addition, with regard to the autocorrelation problem, we first use the DW test. However, the results are inconclusive (critical values: dL= 1.73752< 1.7852< dU= 1.83992; dL= 1.72883< 1.7343< dU= 1.84876), we further use the LM test to examine. The results include non-autocorrelation (0.6839, 0.1121 > 0.05). In addition, we use the Heckman two-stage model (HTSR) to test the self-selection bias, the estimates of the bias parameters (IMR= the inverse Mills ratio) is not statistically significant, and thus its mission would not lead to biased standard errors. We just use the OLS model to estimate the participation model after the robustness test by using the Heckman two-stage sample selection model. Then, after we test the endogeneity with the method proposed in Hausman (1978), we conclude that explanatory variables are endogenous. We use the TSLS model to test the robustness of the endogeneity problem. In addition, we use the LOGIT model to conduct a robustness test. Although the result is not recorded here, the tenor of the results is qualitatively unchanged. Furthermore, tests for endogeneity indicate a potential problem among the control variables. In an effort to solve this problem, lagged values for all control variables are utilized, as suggested by Greene (1997) and
TABLE 4
Relation between Interest Rate Derivative Usage and Risk Exposure
Dependent variable = Interest rate derivative usage Dependent variable = Interest rate risk exposure
Independent variable Excepted sign TOBIT FEVD TSLS VIF OLS FEVD TSLS VIF
Constant - 0.885557*** Interest rate risk exposure + - 0.000328*
( 0.0612 )
Asset-Liability management - - 0.335353
( 0.2445 )
Breusch - Pagan chi-squared– Heteroskedasticity 10.5954 10.3991 11.0031
Notes: We ensure there is no collinear relationship within our analysis (all calculated VIFs are smaller than 10), and we get the result of the regression, including the variance consistent statistic (the Breusch - Pagan chi-squared test = 10.5954, 10.3991, 11.0031) are smaller than χ2 value (12.59159). Besides, we use the DW test to sure the model includes non-autocorrelation (critical values: 1.8421> dU= 1.83992;
2.0229, 2.1397> dU= 1.84876). As for robustness, the results of the LM test suggest that the panel regression is the most appropriate model. Then according to the result of the Hausman test, the model trends to the fixed effect (FE) model. However, we have time-invariant and rarely changing variables within the model. We further use the FEVD technique to eliminate the potential endogeneity bias by Plumper and Troeger (2007, p.129). Then, we find the result of the estimated coefficients is similar for both FE and FEVD models. Besides, due to the extent model uses the censored data, we use the TOBIT model to conduct a robustness test. Furthermore, after we test the endogeneity by the method in Hausman (1978), we conclude that explanatory variables are endogenous. We still use the TSLS model to test the robustness of the endogeneity problem. In addition, tests for endogeneity indicate a potential problem among the control variables. In an effort to solve this problem, lagged values for all control variables are utilized, as suggested by Greene (1997) and Kennedy (1998). The TOBIT, FEVD and OLS method are lags to test within this model. ***, **, and * represent statistical significance at the 0.01, 0.05, and 0.1 levels, respectively.