Hlavní práce75327_buih06.pdf, 1 MB Stáhnout

(1)

University of Economics and Business in Prague Faculty of Finance and Accounting

Finance and Accounting

MASTER THESIS

FACTORS INFLUENCING

REINSURANCE DEMAND IN THE UK

Author: Bui Thanh Huyen Supervisor: Ing. Petra Tisová Academic Year: 2021

(2)

Declaration of Authorship

Hereby, I declare that I compiled this thesis independently, using only the listed resources and literature, and the thesis has not been used to obtain a different or the same degree.

I grant to the University of Economics in Prague permission to reproduce and to distribute copies of this thesis document in whole or in part.

Prague,

________________________

Signature

(3)

Acknowledgments

Hereby, I would like to sincerely express my gratitude to my supervisor Ing. Petra Tisová for her precious suggestions and evaluations throughout this diploma thesis.

(4)

Abstract

This thesis analyzed the relationship between insurer-specific and industry-specific factors to reinsurance demand in the United Kingdom. A dynamic panel data model with Generalized Methods of Moments has been implemented on the data set of 24 representative insurance companies in the UK within the period of 2010-2019. The findings provided evidence that both financial and operational aspects of insurers, pricing and performance of the reinsurance industry are significant determinants of reinsurance decision. On the other hand, there is no substantial evidence for the influence of insurers’

size and leverage usage on reinsurance purchase.

This piece of work has provided a meaningful insight to the UK and global insurance market’s participants and policymakers to better regulate market operation.

Moreover, it delivered a significant contribution to contemporary academic literature.

JEL classification G22, E66

Keywords insurance, reinsurance, insurer-specific, industry-specific, UK market

(5)

List of tables

Table 1: Variable specification ... 14

Table 2: Descriptive statistics ... 30

Table 3: Pair-wise correlation ... 31

Table 4: Dynamic panel regression with one-step GMM ... 33

Table 5: Arrelano-Bond, Sargan and Wald test results ... 34

Table 6: Dynamic panel regression with two-step GMM ... 36

Table 7: Arrelano-Bond, Sargan and Wald test results ... 37

(8)

viii

List of figures

Figure 1: Gross and net Gearing ratio (%, year-end) ... 26 Figure 2: Combined and loss ratio (%, year-end) ... 26

(9)

ix

Acronyms

EIOPA European Insurance and Occupational Pensions Authority IAIS International Association of Insurance Supervisors

GIMAR Global Insurance Market Report GMM Generalized Methods of Moments ROE Return on Equity

UK United Kingdom US United States

(10)

1

1. Introduction

1.1 Background

For centuries, insurance has been used to diminish uncertainty and protect people from losses. Insurance plans are constructed to cover for perils that are insured in insurance policies which could be for vehicles, housing, property, personal casualty and even fatal events of the insured. However, as the insurance industry continues to flourish, there are substantial risks that exceed insurers’ capability. The ideology of “reinsurance” was introduced to share the risk of insurers. Holland (2009) provided a proper definition that

“Re-assurance may be said to be a contract, which the first insurer enters into, in order to relieve himself from those risks which he has incautiously undertaken, by throwing them upon other underwriters, who are called re-assurers". The demand for a distinctive method for huge risk management emerged since the fourteenth century with the evolution of marine insurance (Holland, 2009). According to Gerathewohl (1982), the initial reinsurance agreement dated 12^th July in 1370, was written in Latin and regarded the safety of a cargo ship. The extra hazardous path of the journey was transferred to another more capable insurance party which is alleged now a “reinsurer”. At that time, the reinsurance treaty was still treated as a sales contract, despite in practice trading the risk. Those contracts failed to indicate any premium amount, which was distinctive of insurance regulations in that century, but they already retained certain typical characteristics of reinsurance contracts in the present days. The owners of insured possessions were intact under the reinsured-reinsurer transactions. Reinsurance contracts were independent of the original insured. Reinsurance has long been regarded as the backbone of insurance for efficient capital management and risk control.

(11)

2

What was the next milestone for the reinsurance industry? As England defeated Spain in 1588, England became the world leader in terms of power, resources, trade. Since the Great Fire in 1666, fire, accident, marine and life insurance have become more common in England. During the 18^th and 19^th centuries, the UK achieved the highest level of insurance concentration in Europe, as the result, commercial and personal insurance became the cornerstone for economic and social development in the UK. UK has been the birthplace of many modern insurance types, until now the UK is still the key player in the international insurance market (Haueter, 2013). Being such a robust market for the inception of insurance, reinsurance was blooming with large written volume in the 1800s.

But it was noteworthy to state that in UK domestic market, reinsurance was mostly conducted by insurers, while in Continental Europe, specialized reinsurers were responsible for that. Due to the preference for co-insurance, the UK market was not a competitive landscape compare with Germany and Switzerland. As the consequence, UK’s reinsurance balance was typically dominated by importing, other than exporting.

UK’s insurance market gradually thrived, the need for cooperation of small and medium business and risk diversification brought Lloyd's to life. Rooted from a coffee house in London, Lloyd’s has been transformed from a joint venture project to share the losses of ship owners into an unsettled syndicate of dedicated underwriters aiming for wide-reaching professional businesses (Lloyd's, 2021). Lloyd’s now has an unrivaled assortment of insurance specialists acting under Lloyd’s reputable umbrella and providing services and analysis on a worldwide basis, which become a unique feature when mentioning the UK’s insurance market.

(12)

3 1.2 Purpose of the study

From the publication of the Association of British Insurers (2019), the UK insurance market now landed at the fourth place in the world and first place in Europe in terms of total written premium. As a progressive marketplace nurturing the blueprint of reinsurance and also one of the biggest insurance industries in the world with extensive economic strength, the UK’s insurance market has gained increasing attention from insurance-related entities. In the light of that, not many researchers have delved in-depth into determinants of reinsurance demand in the UK. While several other studies were concerned about reinsurance demand in New Zealand, Sweden, the US and CEE countries (Adams (1996), Gron (1999), Kader, Adams, Andersson, & Lindmark (2010), Curak, Pervan, & Kramaric (2014)), peculiarly there are no papers that target the UK in these recent years. Therefore, the topic: “Factors influencing reinsurance demand in the UK” was chosen as the subject of this paper. This study will seek to examine the necessity for reinsurance based on factors deriving from both insurers and industry.

1.3 Study boundaries

The insurance market has been evolving alongside with economic and social conditions of countries. Insurance demand is stimulated with the development of the economy and society since the population with more stable life care more about their health, retirement income and also have better interest and financial foundation for insuring other properties.

Furthermore, the development of society leads to an increase in great loss and higher demand for corporate indemnification. In general, that is how reinsurance demand originated. Observing the growth of reinsurance premium throughout the years in the International Association of Insurance Supervisors (2019), it is noted that from 2010, the

(13)

4

reinsurance sector started its steady upward trend after a long time of stagnation. So this study’s time frame was circumscribed from 2010 to 2019 for the dynamic nature of observables.

To analyze the impact of insurers and industry on reinsurance purchase, a dynamic panel model with Generalized Methods of Moments (GMM) estimators will be implemented with the 24 biggest insurers in UK territory and some global reinsurance index. This decision stems from the considerable total market share those key players are accounting for as well as the limit in data available. By this, this study aims to observe the effect of insurer-specific and industry-specific factors on reinsurance decisions in the UK in a 10-year period, but not to construct a universal forecasting model for the fluctuation of the reinsurance sector.

1.4 Contribution

Previous research about reinsurance demand already emphasized the influence of insurer-related factors as the predominant contributor. Further studies that also shed light on industry attributes are only Cole & McCullough (2006), Lei & Schmit (2010) in the US market, and (Curak, Pervan, & Kramaric (2014) in Croatia. By analyzing both insurer-specific and industry-specific factors inconclusively in the UK insurance market, this study has differentiated itself in the current literature of the same researching topic. Additionally, on the way to construct some new added value to the field, the author has attempted to apply suggestions of prior researchers and make improvements. This aims to support the accomplishment of previous studies and deliver a complete contribution.

(14)

5

This study does not only contribute to existing research with new and comprehensive findings in a different geographical context but also shows its value for the development of the market operation. When UK’s insurance market continues its expansion, insurers will soon seek for reinsurance to transfer the risks, the outcome of this paper will be of interest. This could be a source of market knowledge for the insurers who are existing in the UK market or inclined to penetrate the UK market, and also the reinsurers to approach more clients. This can benefit market governors for control and renovation market regulations as well.

1.5 Disposition

The structure of this paper consists of six main parts. After the introduction follows the theoretical framework to provide theoretical considerations and review of existing literature to summarize ideas and results of previous scholar performance. Afterward, the research design will continue to specify methodology, describe the regression model, introduce the meaning and description of variables. The empirical result of the model and discussion is illustrated next in the analysis section. The last part will be saved for the conclusion and some suggestions for further research.

(15)

6

2. Theoretical framework

The key question of the study leads to the ultimate micro-economics Demand and supply theory. As the essential groundwork of many empirical studies, comprehending Demand and supply theory is vital when examining what impacts reinsurance needs. As defined in Mankiw (1997), demand is the quantity of goods and services that consumers are willing and able to buy at a given period. Demand is essential to any kind of economy operating and rolling. Consumers demand goods or services to fulfill their wants and needs. How much consumers demand for goods and services also represents their satisfaction that they expect to gain when owning them, which is known as “utility”.

The promising level of utility from the product is the first and foremost determinant of demand. The second driving factor of demand is consumer income. In reality, consumer earnings significantly control their decision to shop, regardless of how much utility the products can bring about.

Besides demand, the trade balance cannot function without suppliers. Quantity supplied is the amount of goods and services that sellers offer and sell to the market at a given market price. Quantity supplied is directly under the impact of price, sellers can always decide to propose more or less of their products to market when the price is suitable or minor for them.

The principle of demand and supply is one of the key concepts to analyze when investigating any industry and product progress. The Law of supply and demand administers the whole market system by governing price. Supply and demand laws stated that in order to obtain the equilibrium price, the supply and demand should equal.

In general, the price of the product is determined by both supply and demand quantity

(16)

7

and vice versa. The product price is positively correlated with the rise in quantity demanded and supplied, while the downturn on both sides will lead to a decrease in the product price. This is the core message that every market participant has to recollect and abide by so as to survive and grow in the competitive marketplace.

Taking the theoretical framework into account, the study’s model was developed with both firm-specific and industry-specific factors. Firm factors are attributed to the consumer side of theory while industry contribution is obviously in the supplier's facet.

Even though in this paper, only demand for reinsurance is scrutinized, the interrelated association between supply, demand opponents and product price rendered it indispensable to research both two parties to comprehend the actual cause for the demand for reinsurance.

When discussing an insurance topic, it is well-meaning to mention Corporate risk management techniques. Similar to other industries, insurance companies have to deal with several kinds of risk in daily business life, an appropriate risk management model is always an effective tool to hedge the risk and sustain an efficient capital structure. For decades, multiple traditional risk control mechanisms by integrating insurance and derivatives have been implemented and become a vital component of insurance company’s risk strategies. Banks (2004) introduced an advanced hedging technique as an alternative risk transfer with the combination of “innovative insurance and capital market solutions”. Among three main approaches for risk management which are “loss control, loss financing, and risk reduction”, loss financing was essentially significant with insurance firms as it deals with large losses and ensures a reasonable amount of funds in the event of risk exposure. Instead of circumventing the

(17)

8

threats, the loss financing technique is more concerned with transferring risk to other entities, for example, a reinsurer.

What reinsurers bring to reinsured is beyond a pure loss financing method, but a rational business strategy to maximize firm value and optimize funding plans. While being a constructive tool to control risk, it is costly to utilize and takes insurers time and effort to consider before making risk-transferring decisions. Lee & Lee (2012) specified that the reinsurance business links with the reinsured’s better risk capacity management, thus enhances the firm’s performance and profitability by dealing with underwriting risks and improve investment activities. The study also added the motivation to reinsurance purchase is the “ability of residual claimants to effectively hedge against operational risk”. Hoerger, Sloan, & Hassan (1990) agreed that insurers are stimulated to buy reinsurance even though it might reduce expected profit margin because reinsurance helps to guarantee the “predictability of cash flow and lower the volatility of earnings”. Since the transformation of the solvency capital regime from Solvency I to Solvency II, reinsurance has captured more interest from insurance companies as an instrument to release the capital burden. In a nutshell, understanding the mechanism of reinsurance as a risk transfer method in the risk management strategy of insurance firms will be beneficial to study the root of reinsurance demand.

(18)

9

3. Literature review

A plethora of previous studies has examined the demand for reinsurance with different methods throughout the years. Regarding Lloyd’s territory, Mayers & Smith (1990) first suggested organizational characteristics directly affect reinsurance decisions of property and casualty insurance corporations. The research emphasized the dependence of ownership structure, firm size and the line-of-business concentration on the reinsurance transactions. The contribution of firm-specific factors was more highlighted in Adams (1996). Analysis which was conducted in the New Zealand life insurance market advocated the impact of firm features, especially the scope and level of leverage. It was stated that reinsurance demand is highly associated with smaller- scale firms with a large leverage ratio. This study is forthright concerned with the risk- bearing hypothesis which was equivalent to risk financing techniques mentioned in the Literature review as the impetus of reinsurance demand.

The financial situation of insurance companies was under discussion additionally in Gron (1999) where more monetary parameters including loss ratio, the proportion of financial obligations were considered to evaluate in detail the probability of financial distress. An innovative contribution of this research could be the mention of liquidity indicators. While other studies referred profusely to leverage as the key stimulus to reinsurance calls, liquidity and solvency took some significant parts too.

Besides observing how highly leveraged firms could be to utilize reinsurance, it is fascinating to realize, highly solvent firms could rely on reinsurance as well; or insurance firms managed to achieve solvency outcomes with the help of reinsurance. A more recent study, Kader, Adams, Andersson, & Lindmark (2010) is furthermore an

(19)

10

advocate for the hypothesis. In addition to the emphasis on the adverse impact of profitability, the study pointed out the positive correlation between liquidity fraction and reinsurance demand. Specifically, in the fire insurance market of Sweden, it was statistically proven that fire insurers adopted reinsurance as a method to relieve the fund burden and back up their incomes. In that way, they are able to accumulate a considerable amount of cash reserves and initiate multiple potential investment opportunities.

In 2003, another study explored further the US property-liability market with a primary focus on the profitability aspect of the insurance firm’s reinsurance decision.

Garven & Joan (2003) had chosen to inspect leverage ratio, investment return, claim cost, diversification of long-tail insurance lines and concentration on tax-favored resources as independent variables. The outcome was aligned with other scholars’

expectations that the more leverage and costs insurance firms are subject to, the higher the usage volume of reinsurance. Meanwhile, the more profitability the firms earn, the less reinsurance they might count on. US’s reinsurance demand analysis was continued with Cole & McCullough (2006) which also specified that financial and operational features of insurance companies consisting of size, group affiliation and organizational form are substantially attributed to the need for risk transfer. In this study, the diversification in the lines of insurance business also was stressed similarly in Wang, Chang, & Tzeng (2008). Especially, Wang, Chang, & Tzeng (2008) argued employing reinsurance services could offer more benefit in terms of business expansion to insurance firms with a more particularly diversified product mix because consequently, they are encouraged to insure more business in different lines. Both of the researchers

(20)

11

have hypothesized the role of the vast business mix could lead to reinsurance incentives as the more spread range of products are, the more risk and uncertainty the insurance firms are exposed to. Reinsurance will be accommodating to provide “specialized knowledge and economies of scale” for multi-line firms.

Since 2006, the pioneering philosophy regarding the role of reinsurance companies and industry facet in reinsurance demand analysis started to capture attention. Cole & McCullough (2006) was the first comprehensive study reflecting those factors. Besides highlighting the importance of financial and operational features of insurance entities, it postulated the costing, liquidity and also reserve management of reinsurance providers. The reinsurance price was significantly proven to spur reinsurance adoption, while there were no statistically explicit indications that loss reserve and liquidity situation could explain for reinsurance decision. The comparable ideology sustained in Schmit & Lei (2008) which focused on reinsurance prerequisite in medical malpractice in the US. Schmit & Lei (2008) has conducted an equivalent model with a slightly alternative variable set. To demonstrate the impact of industry- wide factors, industry price, liquidity and loss reserve fluctuations had been examined.

However, opposite to the author’s anticipation, none of the three reinsurance industry- specific variables have displayed statistical significance to the dependent variable whereas all the organizational factors were verified statistically as a good fit for the model. As approaching more recent academic literature, Curak, Pervan, & Kramaric (2014) featuring reinsurance market in Croatia is an inclusive paradigm with more updated and agreeable outcomes. Supporting the notion of prior researchers, Curak, Pervan, & Kramaric (2014) employed a brief yet exhaustive analysis with both sides of

(21)

12

the reinsurance transaction. By applying a more progressive panel data and method, the industry-specific factors including reinsurance price, financial health of the reinsurance industry were all statistically significant.

In summary, even though a plethora of studies have been conducted regarding the topic, not many have delved into both sides – demand and supply of a product market, especially within the UK territory which is one of the most dynamic and progressive insurance markets in the present days. For that purpose, the intention of this study is evident as a meaningful literature contribution to the current reinsurance analysis and the growth of the UK’s insurance market as a whole and other market participants in specific.

(22)

13

4. Methodology

The purpose of this section is to provide in a thorough manner how the model of this study was implemented to justify the hypothesis and answer the ultimate research questions. The structure of this section will first comprise a description of variables included in the model and an explanation of why they were appropriate for this study. It will be followed along with the estimation of impact to the study’s goal subject – reinsurance demand. Thereafter, the research design is clarified in the subsequent part.

The research design will further specify how the regression model looks like to present the relationship of independent and dependent variables and also certain kinds of auxiliary tests for decent empirical results.

The rest of the methodology will be referred to as how the data collected and modified. Due to the limitation in data available, the sample of this study will constrain to 24 key market participants. The structure is divided into two parts regarding insurance companies and the reinsurance market. How raw statistics were fine-tuned to better conform to the regression parameters is also mentioned. Statistics are mainly imported from companies’ annual financial reports in the period 2010-2019.

4.1 Variable specification

Data support for the model was collected from various sources mainly from the insurer’s annual financial report. Firms’ main official website has provided sufficient statistics regarding their current and historical financial statements. By that informational portals, the study has utilized premium, losses along with other financial indicators belonging to 24 insurance and reinsurance groups of the UK in the period of 2010 to 2019. Figures

(23)

14

were afterward refined to be suitable for the variable’s parameters which are present in detail in Table 1.

Given prior intellectual performances, this study plans to evaluate both industry- wide and insurer-specific factors in the relationship with reinsurance demand, which is anticipated to yield an up-to-date and inclusive empirical contribution for the ongoing researching territory. The explained variable was exhibited by the relative amount of premium insurance firms decided to transfer to reinsurers. Variables are hand-picked from the most reliable indicators used in previous literature. Control variables that represent insurers’ characteristics primarily illustrate the insurer’s financial and organizational situation, which consists of probability to default, capability to handle claims, profitability and investment efficiency. On the other side, industry explanatory variables depict market price, financial outcomes and liquidity.

Table 1: Variable specification

Variable Description Forecasted sign

Dependent variable

Reinsurance demand (ReD)

Premium ceded to reinsurance/ total written premium

Independent variables

Size Natural log of total admitted assets -

(24)

15 Leverage

(Lev)

Gross written premium/ policyholder’s surplus

+

Loss volatility (Loss_Vol)

(𝑙𝑜𝑠𝑠𝑒𝑠 − 𝑙𝑜𝑠𝑠𝑒𝑠 )/𝑙𝑜𝑠𝑠𝑒𝑠 +

Product diversification (HHI)

Line-business concentration Herfindahl Index

+/-

ROE After-tax annual profit/ total equity -

Independent variables

Combined ratio (CombR)

Sum of the loss ratio and expense ratio -

Gearing ratio (GearR)

Recoverables/ total capital -

Source: Author’s compilations

 The traditional approach to reinsurance demand

Size is the most routinely used variable in the academic journey to analyze reinsurance demand. Since the beginning of analyses, Mayers & Smith (1990) has introduced organizational size as one of the most deciding factors that leads to reinsurance reliance.

The size of a firm in terms of financial performance represents the firm capability to handle losses and compete in the marketplace. Bigger firms are anticipated to own more

(25)

16

generous capital reserve, thus better react in risk exposed situations. Furthermore, the size of a firm can act as a proxy to bankruptcy probability, which Hoerger, Sloan, &

Hassan (1990) have declared as a significant influence on corporate demand for insurance. As stated in other researches, size impact is expressed as a natural logarithm of total gross admitted assets to test the correlation with size and reinsurance purchase.

It was expected in many other papers that smaller insurance companies with greater bankruptcy costs will be more inclined to depend on reinsurance to cover their extra business.

Another outstanding factor attributed to bankruptcy risk is the leverage figure.

Opler & Titman (1994) found that highly leveraged firms are more disadvantaged than less leveraged counterparts in crisis time. In industry downturns, firms with higher leverage will have more chances to undergo sales and market share shrinkage than others with less leverage, hence those firms are more susceptible to bankruptcy. It was concluded that an increase in leverage level renders insurance companies to be more vulnerable to financial distress. The rationale is sustained in the insurance field with evidence supported by many prior studies. Once reinsurance is activated, the outcome for insurers is the diminution in cash flow fluctuation and leverage dependence. Garven

& Joan (2003) hypothesized that “reinsurance is essentially a substitute for surplus in terms of leverage effect”. It also suggested expressing leverage as the ratio of written premium over the surplus of the firm. With the evident benefit to relieve the need for leverage tools and solve insolvency dilemma, reinsurance volume is expected to be positively correlated with the leverage index.

(26)

17

Regarding business uncertainty, loss volatility is a remarkable parameter to deliberate. Loss volatility is a meaningful indicator to measure the ability of insurers to manage annual policy compensations. Loss volatility depicts the annual development of incurred loss in relation to direct premium written over the years, as already specified in Hoerger, Sloan, & Hassan (1990) and Curak, Pervan, & Kramaric (2014). The numerator of loss volatility fraction significantly shows the unearned income of insurers in its underwriting activities because the profit was transferred to policyholders for resolving claims. Loss volatility works particularly to evaluate the financial health of insurance companies by assessing their primary expense and profit. Positive loss volatility indicates insurers’ loss is growing over time, which might signify a financial distress alarm. The same idea applies to excessive loss volatility, which means insurers collect less premium compared to the loss they have to cover. Under this logic, the estimated impact of loss volatility is positive to reinsurance demand, since the more loss exposure insurers have to cover, the more reinsurance will be needed.

Apart from bankruptcy cost and loss threat, product diversification is commonly used in previous reinsurance demand analysis. Cole and McCullough (2006) and Adams (1996) have found that the scope of product mix significantly influences the decision for reinsurance. However, the direction of impact was still in question with several assumptions examined. Firms with a variety of product lines or geographical distribution are likely to expose themselves to more risk, hence utilize reinsurance more frequently to mitigate the risk burden. Under that notion, reinsurance demand is positively associated with a diversified product range of insurers. On the other hand, the opposite point of view was raised regarding firms with a more concentrated business

(27)

18

mix. With a contracted line of products, insurers can look for reinsurers to expand their business with various business opportunities in different market areas because reinsurers are equipped with more risk-bearing expertise. Considering this mindset, reinsurance demand is in negative relation with product diversification. The direction of correlation can be tested further by the Herfindahl-Hirschman industry concentration index which is calculated by summing the squares of the market share of each business line (Hirschman (1945) and Herfindahl (1950)). According to Hirschman (1945) and Herfindahl (1950), the Herfindahl-Hirschman Index (HHI) of 1,500 to 2,500 is considered to be a moderately concentrated business line while HHI of equal or over 2,500 is a highly concentrated business mix.

Insurers have made use of reinsurance not to exclusively hedge the various kinds but also to boost profitability, so profitability’s proxy should be considered. Profitability has long been investigated in Mayers & Smith (1990)’s perspective to measure impact on corporate insurance demand. By assessing different consequences resulted from reinsurance decisions, Mayers & Smith (1990) figured out that reinsurance could eliminate the anxiety for possible large losses of firm managers, hence grant insurers more liberal business prospects. The hypothesis focused on the statement that reinsurance could actually solve the under-investment problem. Mayers & Smith (1990) used leverage to confirm the hypothesis while Cole & McCullough (2006) took return on assets, but both of the studies delivered the same result concerning the negative association of profitability and reinsurance purchase. Firms that better perform in profitability will be capable of dealing with unexpected losses and insolvency problems, thus less reinsurance is in need. Considering equity owners are the key members of the

(28)

19

firm operation, this study has chosen return on equity as a proxy to profitability. Return on equity index is formulated as annual after-tax profit over total equity.

 Industrial motives to reinsurance purchase

Besides the contribution of insurer-specific characters, industry indicators are expected to be statistically valuable. As specified in the demand and supply theory, a product demanded quantity is controlled by five main factors including consumer income, price of related goods, buyer tastes, the expectation for future industry escalations and eventually product price, According to fundamental micro-economics rationale in Mankiw (1997), lower price of the product will give a rise to consumer’s interest for the products, this logic should hold for reinsurance demand and supply. Cole &

McCullough (2006) and Curak, Pervan, & Kramaric (2014) had set combined ratio as a proxy for reinsurance price that might administer reinsurance purchase. This study also followed previous literature and depicted combined ratio as a fraction of the sum of losses and expenses over gross written premium. As a representative of pricing, the combined ratio is supposed to have an inverse relationship with reinsurance demand.

The economics fundamental theory applies reasonably in practice because the higher reimbursement and expenses reinsurers have to reconcile will force the price to be raised, thus negatively affects the insurers’ incentive for reinsurance.

Another measure taken into account is industry leverage. Similar to a business entity’s operation, leverage is an important figure to evaluate the trading efficiency of a market. Leverage is usually reflected in the gearing ratio. Gearing ratio referred to how firms deal with payables by shareholder’s fund. When the gearing ratio of the

(29)

20

reinsurance market is high, reinsurers are more vulnerable to economic downturns, thus impacting risk-hedging performance to insurers. Reinsurance gearing ratio equals the number of recoverables divided by total capital. The higher gearing ratio is expected to undermine reinsurance decisions as it is likely to negatively influence insurers’

confidence to the market operation.

4.2 Research design

 Dynamic panel with GMM

With the above-specified variables, a quantitative method has been conducted with linear dynamic panel data modeling. Because the research goal is to observe the interaction of both cross-sectional dimensions and time series contributors, panel data was suggested to be more suitable. In addition, the panel data was chosen since it showed multiple advantages over the cross-sectional and time-series model. Panel data outperformed the other counterparts in terms of better accuracy in model parameters, “better explaining the complexity of observations’ behaviors with much simplified computation and statistical inference” (Hsiao).

Greene (1997) recommended transforming panel data to dynamic type by adding some lags panel data to better observe the fluctuating progression of events. A dynamic regression model is typically characterized by the inclusion of lagged dependent variables with the aim to highlight the dynamic nature of the study’s research subject (Das, 2019). It was additionally stated that employing lagged variables is likely to cause endogeneity problems, therefore the quality of estimators is not guaranteed. The Generalized Method of Moment (GMM) came to application from this perspective. Studies applying GMM estimation for linear dynamic panel model has been growing substantially over time

(30)

21

because of its advantages to deliver an optimal statistical outcome. By the word of Ahn &

Schmidt (1995), GMM could be used to eliminate certain issues including serial correlation, heteroskedasticity and endogeneity. It is especially effective when the sample is limited with time periods but more cross-sections, which is beyond relevant with this study. Subsequently, a linear dynamic panel model with GMM estimators has been conducted in this study with the following equation:

𝑅𝐷_{( , )}= 𝛼 + 𝛽𝑅𝐷_{( ,} ₎+ 𝛾 𝑍 + 𝛾 𝑋 + 𝜀_,

where

𝑅𝐷_{( , )} is the amount of reinsurance that insurers i demanded in year t (with i ranges from 1 to N and t escalates from 1 to T) divided by the total written premium insurers gained in the same accounting year

𝑅𝐷_{( ,} ₎ is the amount of reinsurance the insurers i demanded at one year lagged (t-1)

𝛼 is a constant term

𝛽, 𝛾 , 𝛾 are the coefficients of corresponding attributes 𝜀_, is the idiosyncratic error term

For linear dynamic panel data models with GMM estimators, certain post-tests are critical to delivering valid results. Based on Arellano & Bond (1991)’s proposal, GMM estimators will perform the best in conditions of “no serial correlation in the errors and no strictly exogenous variables”. In order to check the validity of variables, the Sargan test (1958) of overidentifying restriction should be implemented. In the meantime, (AR)1 and (AR)2 Arellano & Bond (1991) test can deal with the first and second-order serial

(31)

22

correlation issue. Eventually, there is necessary to implement a Wald test for joint significance.

 Sargan (1958) test

When the difference and system generalized method of moments (GMM) estimators are increasingly popular, the potential issues affecting the quality of the model have been much of concern. In the case of the length of the panel restricted around 3, GMM will produce one or two instruments per instrumenting variable. The number of instruments certainly increases in relation to the sample size, which is typically worsened in a small sample. Instrument proliferation will result in overfitting endogenous variables and also inaccuracy in the power of the optimal weighting matrix. The demand to conducting the Sargan test (1958) to investigate the validity of instruments is in dire need. The logic of the Sargan test can be simplified as follow:

“Null hypothesis:

Ho: It is accepted that all the overidentifying restrictions are valid.

Criteria of rejection or acceptation:

Prob>chi-square>= 0.05”

The restrictions of overidentification are confirmed when the probability is below 5%, which also means the instruments are not significant. Therefore, it is always better to obtain a probability over 5% to reject the null hypothesis.

 Arellano and Bond (1991) test

In the view of dynamic panel data, the past observations are likely to be correlated with the present recognition of the explanatory variables. Autocorrelation conveys the degree

(32)

23

of resemblance between the given time series and lagged copy of itself in successive time sequences. Autocorrelation can cause many troubles to consistency and bias of estimated variance of regression coefficients. In order to guarantee the validity of the hypothesis testing, the test for autocorrelation is vital. Arellano-Bond (1991) test for autocorrelation is targeted to “the differenced residuals to eliminate autocorrelated and unobserved disturbance terms”. The operation of this test can be interpreted as follow:

“Null hypothesis:

Ho: Autocorrelation doesn’t exist Criteria of rejection:

The probability pr>z is higher than 0.05”

In an effective and valid dynamic panel model with GMM estimators, the first difference is expected to be autocorrelated because “∆µi;t = µi;t -µi;t-1 is supposed to be correlated with ∆µi;t-1 = µi;t-1 -µi;t-2 as they have common µi;t-1 term”

(Roodman, 2004). Therefore, the p-value of the first-order autocorrelation test should be lower than 5%, the existence of serial correlation cannot be rejected. In the meantime, the second-order test is not supposed to deliver an autocorrelation issue, so the probability of AR(2) is not significant at 5% level of significance will be anticipated.

 Wald joint test

The Wald test (also known as Wald Chi-squared test) is a specialized test to confirm the significance of explanatory variables (Agresti, 2012). This test is dedicated to verifying if the chosen independent variables are contributing something valuable to the model by calculating “the weighted distance between the unrestricted estimate and its hypothesized value under the null hypothesis”. From Wald joint test, the model would

(33)

24

be adjusted accordingly to eliminate the meaningless variables. Wald test is highly advantageous over other hypothesis testing including the likelihood-ratio test and Lagrange multiplier because it alleviates much computation burden and still delivers accurate results (Greene, 1997). Under the current dynamic panel model, the Wald test can be conducted on multiple parameters with a single hypothesis. The null hypothesis of the Wald test is the predictors’ regression coefficient is equal to zero, which means their contributions are not beneficial to the meaning of the model. If the p-value obtained is lower than 5%, we reject the null hypothesis and conclude the coefficient of explanatory are different from zero, thus the impact of independent variables are valid.

4.3 Data and sample

Similar to the research design section, the data and sample part will be separated by the insurer’s statistics and reinsurance industry relating figures. Each part will elaborate comprehensively on data specifications and how they were accumulated. To fit in attributes of variables and the regression model, statistics have been updated and refined.

Detailed progress will be mentioned next.

 Insurer data with a focus on the UK market

The UK is having the largest insurance market in Europe and fourth in the globe. Based on Statista’s report, in 2019 there are 911 authorized general insurance companies operating in the UK market. UK insurance market is characterized by a range of international and domestic players. According to the United Kingdom Insurance Report Q4 2019 of Fitch Solutions, 75% of the market share belonged to 24 key market participants. These 24 insurance companies consist of Life, Non-life and also composite businesses. Due to the constraint of data availability, this study will concentrate on these

(34)

25

24 representative key players to represent the whole UK insurance industry.

Nevertheless, this could also be an advantage to ensure the quality of data since those

“blue-chip” market participants’ information are highly standardized and accurate to be distributed to the public. Moreover, Lloyd’s of London is not a company but a marketplace where 99 different groups of insurance and reinsurance firms (known as syndicates) cooperated for the underwriting business (Figure as of year-end 2018). The fact that Lloyd’s is an inclusion of a variety of insurance firms will certify the diversity of the data set. As listed in the Q1 2019 UK insurance market report, leading companies in the market are Prudential, Aviva, Lloyd’s of London, RSA insurance, AXA, Legal and General Group, Zurich insurance group, Allianz group, Scottish Widows Groups, Ageas, Bupa, AIG, Munich Re, Royal London Mutual, Aegon, Canada Life, Direct Life, Just Retirement, Liverpool Victoria, Old Mutual, Pension, Phoenix, Rothesay Life, Standard Life, Swiss Re. From each company and organization’s official website, annual financial reports from 2010 to 2019 of each key player were extracted and examined for statistics.

 Reinsurance data from global report

Since the insurance UK market is highly globalized with considerable reinsurance business with foreign reinsurers, global reinsurance industry data was excerpted for this study. Two reliable sources to offer quality data for global statistics are the Reinsurance market report by Willis Re and the Global insurance market report (GIMAR) by the International Association of Insurance Supervisors (IAIS), both were considered from 2010 to 2019.

(35)

26

Figure 1: Gross and net Gearing ratio (%, year-end)

Figure 2: Combined and loss ratio (%, year-end)

Figures 1 and 2 have illustrated visibly the development of the most reputable metrics to assess the reinsurance industry’s performance. Under the researching scope of this paper, only combined and gearing ratios are analyzed. The difference between gross and net gearing ratios is the subtraction of collateral and offsetting items. For the most accurate result, the net gearing ratio was chosen in the data sample. As shown in Figure 2, gearing ratios undertook a downward trend since 2009, which was presumably

(36)

27

due to the outpouring in reinsurers’ capital base. Since the capital was intensified, the influence of recoverables on reinsurance was relieved. In the meantime, the UK reinsurance market has experienced a gradual upward movement in the combined ratio.

The relentless movement of these two industrial indications might be derived from the volatility in the market’s profitability. These measures are usually highly vulnerable to large payouts, especially after severe natural catastrophes. With the visualized approach, combined and gearing ratios are observed to move in opposing direction, it is fascinating to see how they react in the relationship with the reinsurance demand of insurers.

5. Empirical result and analysis

The main focus of this study is to answer these two research questions: “Are reinsurance demand affected by both insurer-specific and industry-specific attributes in the UK?” and

“How different are those factors’ impact to reinsurance demand?”. To resolve this question, the research design had targeted the relationship between reinsurance decision, financial and structural aspects of insurers and some industry indices. Dynamic panel model with GMM estimators was chosen to perform tests. This section will present the empirical findings of the main model as well as some analysis to expand the result. The outcome of regression and evidence to questionable hypotheses will be specified first. Following are the statistical reflections and quality testing of the model.

5.1 Descriptive statistics

Table 2 shows the descriptive information for all variables mentioned in the model.

Statistical parameters of variables also indicated some insight into variables’

characteristics. The mean of dependent variable ReD lingered at 0.1515, which means the average premium ceded to reinsurers is around 15% of the total written premium. In

(37)

28

previous studies including Curak, Pervan, & Kramaric (2014) and Cole & McCullough (2006), the result shared a relatively smaller figure. The reason behind this could be the disparity in the market. UK market is characterized by substantial insurance giant groups like Lloyd’s, Aviva and Prudential, which made up for a considerable market share in Europe, which might lead to a higher surge in the greater proportion of premium ceded.

This also explained for the utmost bound of ReD is 80% much higher than results in other markets.

Another indicator noted to mention is Loss_Vol with a mean value of 0.1260. The average growth of claims in UK insurers is around 12.6%. The size of insurers ranged from 7.4 to 13.74, which designates a wide gap between the financial capability of cross- sectional units in the sample. The other studies also produced the same huge disparity in the size of insurers. A recent report from European Insurance and Occupational Pensions Authority (EIOPA, 2019) showed that the insurance industry in the UK and the rest of the world are undergoing a density problem with a certain number of enormous insurers dominating the whole market. The variety in the sample was additionally illustrated in HHI and Lev’s statistical results. As HHI is presenting the diverse business mix of insurers, it also shows different types of risk insurers expose to and different sources of income contributed to insurers’ budget. The mean HHI of observations is 3.690 that is above the typical figure of a normally concentrated marketplace. It is interesting to witness the upper bound of HHI reached peaking 8562 while the lower bound halted at 485 only. It can be inferred to the variance in the business strategy of UK insurance firms. This is within the author’s expectation since the UK is considered a highly dynamic and competitive market with the participation of top-notch insurers worldwide. Lev’s figure was even more

(38)

29

extreme, which oscillated from 0.03 to 3.62. As discussed in the previous section, lower Lev is more advantaged for the firms since it displays the better ability to deal with potential claims by available capital. 3.62 is certainly not a good indication for underwriting operations that might induce a higher demand for reinsurance.

In addition, a brief description of the reinsurance industry was conveyed by CombR and GearR. Comb R representing the Combined ratio, a typical underwriting profitability measure of the insurance industry. The combined ratio is calculated by dividing the sum of incurred losses and expenses divided by the total gained premium.

Based on the result in table 2, the average Combined index of the reinsurance industry within the 2010-2019 period is 65%, which indicates the industry is operating efficiently and making a positive profit. The range between the minimum and maximum extreme is relatively small at 1.5%, that means there is not a big disparity in different moments in the period and different individual company. Alongside CombR, the Gearing ratio is also showing an optimistic picture of the reinsurance industry. The optimal gearing ratio is settled between 25% and 50%, therefore the average result of 47% is a good indicator.

This expressed that reinsures are using a sufficient amount of loans to pay for operational expenses and not expose themselves to elevated uncertainty in economic difficulties.

However, the minimum figure of 38% is comparatively a conservative ratio. Maintaining low leverage is not always a good sign since reinsurers have to rely on more shareholder's equity and possibly not invest enough in economic development.

(39)

30 Table 2: Descriptive statistics

Variable Mean SD Median Min Max

ReD 0.1515 0.1349 0.1200 0.004300 0.8000

Size 11.25 1.900 11.59 7.460 13.74

Lev 1.547 0.7524 1.430 0.0300 3.620

Loss_Vol 0.1260 0.6762 0.03500 -0.8100 4.800

ROE 0.1049 0.08634 0.1000 -0.3000 0.300

HHI 3690 1799 3804 485.0 8526

CombR 0.6590 0.04818 0.6450 0.6000 0.7500

GearR 0.4710 0.07695 0.4700 0.3800 0.5800

Another insight to variables’ statistics provided is a pairwise correlation. Pairwise correlations uncover a potential association between variables that is worth further examination. It helps us detect signs of correlation and possible multicollinearity which may sabotage the precision of estimate coefficients and reduce the statistical quality of the regression model. As stated in Gujarati (2004), a severe multicollinearity problem might be present with the absolute value of correlation larger than 0.7. Under that rule of thumb, there is no evidence for multicollinearity in this regression model. In general, independent variables have shown relatively week relations with each other. Only the size is considerably correlated with the rest, but it did not reach to warning level of multicollinearity. The most noticeable correlation between independent variables is between Size and HHI. As expected, CombR and GearR displayed a minor correlation with other independent variables since in practice, the reinsurance industry hardly has any

(40)

31

significant impact on the insurance firm’s indicators. Regarding the relationship of dependent and independent variables, the majority of variables are negatively associated with reinsurance purchase. Based on the data, product diversification imposed the largest negative influence to reinsure demand. The indicative result is contradicting with the author’s anticipation. However, it is a good sign that the outcome did not disclose any trivial correlations and no multicollinearity issue is revealed. A regression model shall be performed next to further analyze the subject.

Table 3: Pair-wise correlation

Variables ReD Size Lev Loss_Vol

ReD 1 -0.0494 -0.0087 -0.1035

Size -0.0494 1 -0.2262 0.0129

Lev -0.0087 -0.2262 1 0.0190

Loss_Vol -0.1035 0.0129 0.0190 1

ROE -0.0146 -0.1746 0.1658 0.0518

HHI -0.0931 0.4197 0.1547 0.0650

CombR -0.1462 0.0147 -0.0083 -0.0298

GearR -0.0996 -0.0485 0.0840 -0.0056

(41)

32

Variables ROE HHI CombR GearR

ReD -0.0146 -0.0931 -0.1462 -0.0996

Size -0.1746 0.4197 0.0147 -0.0485

Lev 0.1658 0.1547 -0.0083 0.0840

Loss_Vol 0.0518 0.0650 -0.0298 -0.0056

ROE 1 0.0706 0.0898 0.1591

HHI 0.0706 1 0.0570 -0.0257

CombR 0.0898 0.0570 1 0.2251

GearR 0.1591 -0.0257 0.2251 1

5.2 Regression result

This chapter is aimed to interpret the regression results of the relationship between the firm-specific and industry effects and reinsurance usage in the UK’s insurance market. It is going to analyze the impact of both operational and financial factors of insurers including size, leverage, loss volatility, product diversification as well as performance and pricing of the whole reinsurance industry to corporate motivation for reinsurance in 24 UK insurance companies from 2010 to 2019. Because of the overwhelming diversity in insurers’ data set of the UK market, 24 insurers with top performance in both life and non-life segments were

(42)

33

selected as the sample. The ranking was provided by the United Kingdom Insurance Report in Q4 2019 of Fitch Solutions. The difference in insurers’ financial situation, loss capability and product portfolio are potential determinants of the disparity in reinsurance necessity.

According to the fundamental microeconomic theory, the role of industry pricing and performance should also contribute some impact. Both of these two hypotheses will be verified by the dynamic panel regression with the GMM estimator.

To initiate regression analysis, the one-step difference GMM dynamic panel model was conducted as suggested in other researches. Under the one-step model, the results were relatively significant at a 10% level of significance; however, in the view of the GMM dynamic panel model, the one-step estimation did not satisfy a certain condition for the model validity. The result was presented in Tables 4 and 5 below.

Table 4: Dynamic panel regression with one-step GMM

Variable Estimate Probability

Intercept 0.0289884 0.0778*

ReD(-1) −0.312629 1.18e-07 ***

Size −0.00320623 0.9735

Lev −0.000895170 0.9719

Loss_Vol −0.0380069 0.0570

ROE −0.0696254 0.3392

(43)

34

HHI −8.86149e-06 0.8475

CombR −2.38931 0.0002

GearR 0.0536590 0.3645

Table 5: Arrelano-Bond, Sargan and Wald test results

Test for AR(1) errors, p-value 0.5450

Sargan over-identification test, Chi-square(35) 0.0000

Wald (joint) test: Chi-square(8) 0.0000

According to Labra & Torrecillas (2018), the two typical problems in the estimation of dynamic panel data with GMM were the outnumber of instruments and serial autocorrelation of errors. The effects can be exacerbated with a sample of a long period of time and a limited number of observes, which is highly likely to be the issue of this study. Along with the dynamic panel model, Sargan (1958), Arellano and Bond (1991) test and Wald test were performed but the outcomes were not promising. Table 5 presented the probable outcomes of these tests. It can be referred from the results that the p-value of the first-order autocorrelation is bigger than 5% while the counterpart of second-order AR(2) is lower. Under this circumstance, AR(1) null hypothesis for autocorrelation was

(44)

35

rejected while the statement in AR(2) was not rejected, which signified the existence of serial correlation in second-order, instead of first-order. As mentioned in conditions to ensure the validity of dynamic panel data, there should be autocorrelation in only AR(1).

The reaping results were against the criteria and required to be accustomed. The demand for amending the regression model was further emphasized in Sargan and Wald test’s outcome. Sargan test’s probability of 0.000 has failed to verify the quality of the model by bearing the indicator of an over-identifying issue. In the meantime, the Wald test has delivered statistically significant statistics with a p-value smaller than 0.05. That means the current explanatory variables are contributing significant influence to the analysis model. The drawbacks of this statistical outcome of one-step GMM estimators have motivated the author to attempt two-step GMM testing.

Under the conventional ideology, both one-step and two-step GMM estimators are asymptotically normal in a large sample. The performance of one-step and two-step counterparts is evaluated with the same efficiency. However, the conservative theory has disregarded the estimation uncertainty in the weighting matrix, thus not inclusively conveyed the case of limited samples. In Hwang & Sun (2015), by using fixed-smoothing asymptotic theory, it was successfully verified that a model with two-step GMM estimators outperformed the one-step process. Additionally, Roodman (2009) dictated that the “two-step estimator is asymptotically eﬃcient and robust to whatever patterns of heteroskedasticity and cross-correlation the sandwich covariance estimator models”.

Under these considerations, two-step GMM estimations of dynamic panel data were conducted with the same sample size and set of variables. Table 6 has summarized the updated statistics produced by a new model.

(45)

36

Table 6: Dynamic panel regression with two-step GMM

Variable Estimate Probability

Intercept 0.0304052 2.44e-035***

ReD(-1) −0.312013 0.0000***

Size −0.0127535 0.5341

Lev −0.00249653 0.3262

Loss_Vol −0.0377668 0.0000***

ROE −0.0738560 1.79e-016***

HHI −1.08596e-05 0.0497**

CombR −2.32269 1.23e-237***

GearR −0.195771 8.01e-05 ***

(46)

37

Table 7: Arrelano-Bond, Sargan and Wald test results

Sargan over-identification test, Chi-square(35) 0.9576

Wald (joint) test: Chi-square(8) 0.0000

Indeed, the result of two-step GMM estimators presented more statistically significant results. This was validated by more meaningful statistics in table 7. From table 7, all requirements for validity of dynamic panel data were fulfilled. AR(1) with a probability lower than 0.05 and AR(2) with a probability greater than 0.05 have satisfied the conditions regarding the autocorrelation issue. The better quality of the panel model also displays in the better Sargan test. Sargan's test at p-value = 0.9576 suggested that we can reject the null hypothesis. Therefore, there is no clue of the negative impact of instrument proliferation and overidentification problem. Furthermore, the Wald test still obtained an expressive result with p-value = 0.0000, which again indicates the meaningful impact of explanatory variables on the whole model.

While the power of this model has been validated with the above tests, further analysis concerning the relationship between dependent and independent variables can proceed. It can be inferred from table 6 that within a 1% level of significance almost all explanatory variables including both insurers-related and industry indices are statically

(47)

38

significant with a p-value < 0.01 while within a 5% level of significance Herfindahl business line concentration is significant with p-value = 0.0497. Nevertheless, evidence for Size and Leverage factors’ significance could not be found. The research shared the same evidence with Curak, Pervan, & Kramaric (2014) about the insignificance of insurers' size.

Still, this result is not expected by the author since Size and Leverage’s contribution to the reinsurance demand has been supported by several previous pieces of research. For example, Mayers & Smith (1990), Lei & Schmit (2010), Curak, Utrobicic, & Kovac (2014), all provided evidence about the significance of corporate size and leverage management in the examination of reinsurance demand in the US. It is also endorsed by a practical ideology that size and leverage are the key drivers of reinsurance demand.

Company size reflects its financial strength and ability to cope with bankruptcy costs.

Leverage refers to higher risk in a company’s insolvency and difficulty in resolving liabilities. Even though, the influence of these two parameters might be different, both of these two features are of paramount importance to insurers’ decision to rely on another source to cede risk.

The result of significance testing implied an impactful relationship between reinsurance demand and both insurer-specific determinants like loss volatility, product diversification, Return on Earnings as well as industry-specific inputs such as Combined and Gearing ratio. Overall, it can be concluded that insurance firms’ loss volume, product portfolio and after-tax profitability have some control over the firm’s need for risk ceding.

Moreover, the regression result has contributed striking findings regarding the influence of market indicators in the UK market while several preceding pieces of research have not been able to provide sufficient proof of the reinsurance industry’s impact on insurers’

(48)

39

decisions. One example is Lei & Schmit (2010) could not find significance in reinsurance industry variables in both of their versions of models. However, at the present, it can be accepted that reinsurance demand is affected to some extent by the price and performance of the reinsurance market too.

Considering the sign of correlation of variables, some interesting points are noteworthy to mention. It is intriguing to observe that all of the coefficients of significant variables are negative. Most of the explanatory variables’ impact on dependent variables are in agreement with the author’s forecast at the beginning of the paper. In specific, the opposing effect of Loss volatility, ROE, product diversification, Combined ratio and Gearing ratio to reinsurance demand is inferred by the negative coefficient of variables.

Based on the research result, Product diversification is negatively correlated with reinsurance purchases. In the section of Variable Description, the author predicted the array of product lines might have a positive and negative influence on reinsurance. Especially in the UK’s insurance market, the product portfolio extremely varied from firm to firm. Small to medium Britain insurance companies always try to own at least three range of products at hand while larger-scale firms maintain around. Multiple insurance firms in the UK are composite of Life and Non-life, which means they offer products deriving from both segments. Offering more kind of products and services will create more opportunities to satisfy more insurance need from different customers and expand customers’ base, but it always leads to more independent financial control. With the negative coefficients of the Product Diversification variable, it can be concluded that insurance firms with larger product coverage will be less dependent on reinsurance. Meanwhile smaller firms

(49)

40

regarding business offerings will be more motivated to use reinsurance to boost their sales and profitability.

Another factor that matches the author’s expectation is ROE. The association between ROE and Reinsurance demand is relatively small at p-value = 1.79e-016 compared to other indicators. This is consistent with the underinvestment hypothesis that insurers who undergo too much-leveraged conundrum will forego valuable investment opportunities, thus need to rely on reinsurance for potential large losses (Myers, 1977).

Besides that, return on Equity (ROE) evaluates the profitability of a company in relation to shareholder’s equity, in another way it depicts the efficiency in working capital utilization of the firms to achieve profits. The author forecasted the opposing relationship between ROE and reinsurance demand under the notion that better capital management will result in more sovereign funding. The dynamic panel regression has shared the same result; thus, the verdict is that reinsurance demand will be less robust with higher ROE.

Among the industry-specific factors, the Gearing ratio is another variable obtaining regression outcome in line with the author’s estimation. The gearing ratio is a reputable metric to compare the owner’s equity and long-term liability. The gearing ratio of all reinsurance firms is accumulated to unveil the leverage situation of the whole industry. A higher gearing ratio suggests the companies in the reinsurance industry are making much use of liabilities. GearR’s coefficient sign −0.195771 is consistent with the author’s forecast.

From this result, the reinsurance demand of insurance firms is expected to increase when the leverage indicator of the reinsurance industry is low. This suggested the attitude of insurers towards a better-performed industry. When insurers believe in an industry with less debt and better operation, they will invest more in reinsurance.

(50)

41

Moreover, the result followed the application of the Microeconomics theory about Supply and Demand and author’s prediction. The Combined ratio (CombR) is statistically significant and negatively correlated with reinsurance volume. Since the Combined ratio consists of the loss ratio and expenses of reinsurers, the Combined ratio is a typical representative of reinsurance pricing. The surge in the price of reinsurance due to inflating expenses and claim compensation will result in the hesitation of insurers to purchase reinsurance.

It is interesting to notice that they are all negatively correlated with reinsurance. It is not consistent with the author’s estimation and also the result of the preceding literature.

It is ambiguous to attain evidence for the negative relationship between loss volatility and reinsurance since the hypothesis was developing in contradicting scenarios. While Hoerger, Sloan, & Hassan (1990) and Curak, Pervan, & Kramaric (2014) found loss volatility is statistically significant to the reinsurance purchase, Curak, Utrobicic, & Kovac (2014) in the Croatian insurance market did not produce any proper relations. The growth in claims annually will render insurers to utilize reinsurance to protect themselves.

Especially in UK’s insurance market, the risk coverage is diverse and there is a remarkable possibility for a large loss, the loss development should move in the same direction with reinsurance. The contradiction between this paper’s result and other research could be explained by the sample size. According to Blundell, Bond, & Windmeijer (2012), GMM estimators in dynamic panel data will perform the best in large sample size, so the sample size of 24 insurers in the UK market will certainly a limitation to this research.