Essays on corporate risk management and insurance

Abstract

This thesis is composed by three different research papers tackling topics in the economics of risk, with particular emphasis on corporate risk management and insurance economics. The overall contribution is both theoretical and empirical, since it provides new methodologies to price and detect insurance risks, as well as offering new insights into the drivers of corporate risk management decisions. The first chapter is a theoretical work focusing on dynamic adverse selection in life insurance. It aims to provide a new framework that can be used by practitioners, as well as researchers, for pricing, reserving and statistical testing of dynamic adverse selection in life insurance contracts. Here, dynamic adverse selection is defined as the situation for which ‘good’ risks are more probable to terminate their insurance policy than ‘bad’ risks. To give an example, consider a policy which pays a benefit only if policyholders are alive at contract maturity. Insureds with worsening health conditions (‘good’ risks) might decide to stop paying premiums, as they will not likely benefit from the contract. For the opposite reason, those with improvements in their life expectancy (‘bad’ risks) will be more inclined to keep paying them. This will increase over time the aggregated risk for insurers. The phenomenon of asymmetric information in life insurance has been extensively studied in the literature. Nevertheless, results regarding its presence are mixed between those who find evidence of adverse selection (see, e.g., Finkelstein and Poterba (2004) and He (2009)) and those who find symmetric information (see, e.g., Cawley and Philipson (1996) and Hendel and Lizzeri (2003)). However, the previous literature mainly focused on adverse selection at inception. Furthermore, they typically tested simple life policies such as annuities or term insurance, and often disregarded many of the different contract features. The chapter proposes a model of selective withdrawals, driven by exogenous and endogenous factors, which links the dynamics of policy termination to contract design. Via this model, the mortality risk profile of policyholders can be represented in terms of an endogenous frailty process, shaped by the relative attractiveness of different contract benefits in different states of the world. This method does not require the knowledge of individual insureds’ health conditions, but only their exit times and reasons. The chapter illustrates, with numerical applications on both simple policies (endowments) and more complex contracts (variable annuities), how this approach outperforms traditional ones in both pricing and testing. The second chapter is an empirical analysis of corporate risk management demand in the reinsurance sector. The work tries to shed some lights into what determines the amount of risk management, or more precisely reinsurance, purchased by insurance companies. In particular, it seeks to answer the two following questions: (1) do more (or less) financially constrained insurers buy more reinsurance; and (2) what is the ‘quality’ of reinsurance bought by insurers with high (low) financial constraints? The current state of the literature is split between who argue that more financially constrained firms purchase more risk management, as it reduces their need of accessing costly external capital (see Froot et al. (1993), FSS), and who on the contrary claim that, since risk management is costly, only less financially constrained companies can afford to engage in it (see Rampini et al. (2014), RSV). Furthermore, the empirical evidence on this topic is also mixed between those supporting FSS’s predictions (see, e.g., G´eczy et al. (1997) and Aunon-Nerin and Ehling (2008)), and those in favour of RSV’s ones (see, e.g., Nance et al. (1993) and Carter et al. (2006)). Using a large and granular panel dataset of reinsurance transactions by US Property & Casualty (P&C) insurers, the chapter attempts to test these two hypothesis in the reinsurance sector by examining the determinants for both risk management demand (as measured by reinsurance coverage) and quality of hedging instruments (as proxied by reinsurer’s counterparty risk). The determinants of reinsurance demand have been widely discussed in the literature (see, e.g., Mayers and Smith Jr (1990), Cole and McCullough (2006), Cummins et al. (2008) and Garven and Lamm-Tennant (2011)). However, to the best of our knowledge, no one before explicitly assessed the impact of financial constraints on demand and, furthermore, no one has evaluated the demand for the quality of reinsurance. The results presented in the chapter show that more financially constrained insurers purchase less reinsurance. However, financial constraints are positively correlated with the demand for more credit worthy and hence expensive reinsurance. This result changes when considering only small insurers who prefer cheaper reinsurance as they get more constrained. In terms of insurer’s credit rating, higher rated insurers tend to prefer higher rated reinsurance. Conversely, lower quality insurers seek (or can afford) lower rated reinsurance. These findings are in line with RSV’s predictions, showing that the costs of reinsurance may outweigh its benefits in the presence of financial constraints. The third chapter offers an empirical overview of corporate risk management decisions involving the designing of the Close Protection security layer (i.e. the range of security measures under direct control of an organization). In particular, the research tries: (1) to understand whether companies operating in conflict areas take into account the investment of counterinsurgency programs when designing their Close Protection layer; and (2) to quantify the economic gains from implementing security measures based on the company’s actual risk exposure. Indeed, there are doubts in the literature on whether these programs actually work (see, e.g., Berrebi and Olmstead (2011) and Berman et al. (2011)). The research design is as follows: (i) spatio-temporal marked point processes (see, e.g., Ogata (1998) and Mohler et al. (2012)) are estimated on a granular dataset of attacks carried out in Iraq during the period 2007-15, to better understand the dynamics of attack occurrences and severities; then, (ii) survey data from the top five security providers in Iraq are used to create Close Protection security benchmarks based on ‘unconditional’ information (e.g., as in OSAC (2016)), as well as ‘conditional’ on the results from the predictive models of point (i); finally, (iii) the economic gains of using predictive models are quantified by computing cost deviations of implementing the ‘unconditional’ security posture vs. the ‘conditional’ one vs. the observed one. As an application, the chapter analyses and compares four different areas presenting different socio-economic characteristics and patterns of attack occurrences. It then discusses in detail a case study based on a medium-sized oil field in the Basra province. The empirical evidence shows that firms on average design their Close Protection based on unconditional information, and sometimes overreacting to spikes in conflict activity that have limited bearing for the exposure at stake. Consequently, in areas with extensive counterinsurgency investments (e.g., Basra province), an appropriate use of spatio-temporal information can deliver average security cost savings of around 30% relative to the ‘unconditional’ benchmark, and of around 50% with respect to security postures driven by overreaction. Instead, areas featuring significant spatio-temporal clustering of events (e.g., Al-Anbar and Baghdad [Red Zone]) may require up to 50% more investment in mobile security.