**Sample Construction**

Due to the lack of an adequate database or statistical survey of derivative activities within the insurance sectors, we use the year books and annual financial reports provided by the Insurance Institute of the Republic of China as our main source of data on derivative use and firm characteristics of insurance companies. We collect annual data for both life and non-life insurance firms from 2001 to 2003 inclusive. During our sample construction process, we found that disclosure concerning derivative activities was not standardized among insurance companies. For example, some subsidiaries or branches of foreign insurance conglomerates provide simplified derivatives-use information over the period of observation, which does not allow us to gather the required data for further analysis. Therefore, those insurers with incomplete information were contacted by telephone or email so as to ascertain the derivative-use variables. Since insurers in Taiwan can only use derivatives for hedging purposes, insurers clearly disclose the use of derivatives for hedging purposes in their financial statements, which makes our study on hedging behavior feasible. Table I shows the number of insurers operating during the period in question and the number of insurers included in our sample. Over 90 percent of insurers in business and all relatively large insurers (in terms of gross premiums written and total assets) are in the sample, so the sample is representative of the insurance industry.

**Methodology**

A number of firm characteristics are hypothesized as having an impact on derivative use. A list of these variables and their definitions are given in Table II. We then employ a model to investigate the use of derivatives among Taiwanese institutions in which derivative activities are formulated as the participation decision and the participation extent. In the analysis of the participation decision, a company is regarded as a derivative user if the year-end derivatives balance in its financial statements is not zero. Based on the binary nature (0 for derivatives nonusers and 1 for users), probit regression is utilized to analyze the participation decision. Various approaches to the participation extent model have been proposed in prior research. The OLS model with Heckmanâ€™s two-step estimation procedure suggested by Greene for correcting heteroscedasticity and inconsistency is applied in examining the magnitude of usage in insurance-related studies such as Colquitt and Hoyt, Hardwick and Adams and De Ceuster et al.. The generalized tobit model proposed by Cragg is employed by Cummins, Phillips, and Smith in studying the level of use among U.S. life and non-life insurers and by Shu and Chen in analyzing derivative use among firms listed on the Taiwan Stock Exchange. Sinkey and Carter argue that a tobit regression developed by Tobin is suitable for analyzing the extent of derivative use. In fact, a tobit regression can be regarded as an alternative to the Heckman two-step estimation procedure. Following Colquitt and Hoyt, Hardwick and Adams and De Ceuster et al., we attempted to estimate the extent of derivative use equation using the Heckman two-step estimation procedure. However, the inverse Mills ratio obtained from the probit models was not found to be significant in any of the extent models. Thus, the extent models were estimated using ordinary least squares, fixed/ random-effects models. this

**Table- 2. Variables and Their Definitions**

Variable | Definition |

Dependentvariable | |

Participationdecision | Participation dummy variable taking the value one for participants, zero otherwise. |

Extent decision | End-of-year derivative positions scaled by total assets |

Independentvariable | |

Firm size | Natural logarithm of total assets |

Financial distress | Sum of total insurance claims and associated interests dividedby earningsbefore interest and tax |

Interest rate risk asset | Mismatch where non-current assets outweigh non-current liabilities scaled by total assets |

Interest rate risk liability | Mismatch where non-current liabilities outweigh non-current assets scaled by total assets |

Growthopportunities | Cash reinvestment ratio multiplied by return on equity |

Reinsurance | Ratio of reinsurance ceded to the sum of direct premiums written and reinsurance assumed |

Tax | The amount of tax loss carry forward scaled by net income |

Institutionalownership | Shareholding proportion of institutional shareholders |

Managerialownership | Managerial shareholdings as a percentage of all shares outstanding |

FX | Total amount of foreign investment scaled by total assets |

NIM | Net interest margin scaled by net income |

INV | Value of underlying assets represented by investment-linked insurance deflated by gross written premiums [for life sector] |

CR | Current ratio |

DOMESTIC | 1 if domestic insurer, 0 otherwise |

HERFC | Herfindahl index reflecting concentration of claim payments |

IR | Net Investment income divided by total assets [for life sector] |

Sample Construction Due to the lack of an adequate database or statistical survey of derivative activities within the insurance sectors, we use the year books and annual financial reports provided by the Insurance Institute of the Republic of China as our main source of data on derivative use and firm characteristics of insurance companies. We collect annual data for both life and non-life insurance firms from 2001 to 2003 inclusive. During our sample construction process, we found that disclosure concerning derivative activities was not standardized among insurance companies. For example, some subsidiaries or branches of foreign insurance conglomerates provide simplified derivatives-use information over the period of observation, which does not allow us to gather the required data for further analysis.