• No results found

Private Trade Credit Insurance: the impact of a loss shock in the credit insurance market

N/A
N/A
Protected

Academic year: 2021

Share "Private Trade Credit Insurance: the impact of a loss shock in the credit insurance market"

Copied!
55
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

1 University of Amsterdam

Faculty of Business and Economics

Private Trade Credit Insurance: The impact

of a loss shock in the credit insurance market

Evidence from the Global Financial Crisis

Master Thesis

Author: Silvia Berberi Date: August 2018 Student no.: 11925876 Supervisor: John Lorié

(2)

2

Statement of Originality

This document is written by Silvia Berberi who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Contents

1. Abstract ... 4

2. Introduction ... 5

3. Theory and Background ... 8

3.1 Trade Finance ... 8

3. 2 Trade Credit Insurance and its mechanism ... 10

3.3 Claims ... 12

3.4 Premium ... 14

3.5 Underwriting Measures ... 14

3.6 Loss Shock Theory ... 15

4. Literature Review ... 17

5. Empirical strategy and Data ... 20

5.1 Methodology ... 20

5.1.1 Extensions to the benchmark model ... 22

5.2 Data ... 23

5.3 Potential Econometric Issues ... 25

6. Benchmark Results ... 27

6.1 Main Regressions ... 27

6.2 Lagged Values ... 32

6.3 Further analysis Global Financial Crisis (2008-2009) ... 34

8. Conclusion ... 44

References ... 47

(4)

4

1. Abstract

Private Credit Insurance is an important part of trade finance that enables trade and protects exporters from the risk of nonpayment. During the Global Financial Crisis, the importance of credit insurance became more evident. Since there is little research and data in this field, there is a need for further analysis about the influence of loss shocks in the credit insurance market. This paper focuses on one of the biggest global private credit insurer and examines how an increase in claims paid during the period 2000 to 2015, affects premiums and the shares of export insured.

Making use of a fixed effects model, this research indicates that the effect of an increase in claims ratio on insured exports between an exporter and importer, results on average in a decrease of 11% of the share of export insured and a rise in the level of premiums of 3.6%. Furthermore, I test the size of the shock before and after the crisis. The results show that during the crisis the impact of the loss shock was higher. These results add to the trade finance literature to better grasp the implication of potential credit constraints in times of financial crisis.

(5)

5

2. Introduction

In today’s world where globalization and international trade are growing at a high pace, managing risk and offering protection to cross-border trade parties becomes increasingly important. The occurrence of various risk factors in international trade that can hinder trade volumes, has given more emphasis to trade finance. Trade finance is composed of products and services that cover the risks involved in the trades. The market offers products such as Letters of Credits and Guarantees, as well as trade credit insurance to facilitate cross-border transactions. Trade credit insurance is an insurance that provides coverage on credit risk on trade receivables; manages the risk of delaying payments or failure to pay. This is mainly done by credit agencies (which are either partly or fully government-sponsored) or private insurers, which are reported to be involved in 11% of all international transactions (ICC, 2016).

There was a rise in private credit insurers in the 1990s in Europe, when the European Commission took the decision to prohibit Export Credit Agencies from issuing insurance for short-term export credit risks (maturity less than 2 years). This enabled credit insurers to gain ground and become the main players in providing short-term insurance. The three main insurers are Atradius, Euler Hermes and Coface with a share of 81.5% of the global market, generating EUR 6.0 billion in premium (ICISA, 2014).

The global crisis by the end of 2008, led to a peak in the claim ratios as high as 111.64 % and a decrease in the share of export insured by private trade credit insurance. The coverage limits were reduced, and prices were increased. There is no substantial research that indicates exact

(6)

6 numbers on these changes, but Berne Union 1and ICC2 stated that the exposure of private credit insurance on exports decreased by 16 percentage points through the period of end 2008 and beginning 2009 ( Van der Veer, 2011). Van der Veer (2015) states that the decrease of credit insurance provided during the Global Crisis led to a decrease in exports of 10-20% in Europe. Also, Auboin and Engemann (2014) state that the reduction in supply of credit insurance can explain around 11% of trade collapse in the financial crisis.

Due to the scarcity of data in the field of trade credit insurance, developments in this industry have not been analyzed in depth. Credit insurance plays a crucial role in international trade. Van der Veer (2015) investigates the effect of private credit insurance on international trade, and concludes that 1 euro of insured export generates 1.3 euros in exports. Given the profound impact of credit insurance on the export industry, it is imperative to understand the dynamics and interplay between credit insurance providers, and the share of exports insured. Dynamics such as loss shocks and the impact this has on the quantity of insurance supplied as well as premiums charged by “the Big 3” are the main areas of investigation in this paper. This has already been tested for one of the “Big 3” global credit insurers (for the period of 1992 – 2006) by Van Der Veer (2017). I add to the literature by exploring further into this topic for one other big credit insurer and examiningto what extent the results differ from Van der Veer’s research. Furthermore, the data in this paper is from 2000 to 2015, which includes the Global Crisis of 2008 and 2009. Thus, I delve deeper into loss shock effects by analyzing the impact of claims during financial crisis on the share of export insured and premium.

Theoretical frameworks around loss shocks and export insurance postulate that an unexpected increase in the level of claims can lead to a reduction in the supply of credit insurance as well as an increase in premiums charged (Van der Veer, 2017). I will be testing this hypothesis by investigating the effect of claims on the share of export insured versus exports, and on the price set by the credit insurer on the customer (premium). In addition to this, it is important to

1

An international not-for-profit trade association, representing the global export credit and investment insurance industry.

2

(7)

7 consider the influence of economic cycles on trade credit insurance providers and the extent to which loss shocks (claims), impact during these periods impact them.

The main research questions I will be investigating are the following:

1. What is the effect of an unexpected increase in claims (loss shocks) on premiums?

2. What is the effect of an unexpected increase in claims (loss shocks) on the share of

exports insured?

3. To what extent do periods of financial crisis influence the effect of loss shocks on the

share of exports insured / premiums?

Questions (1) and (2) draw from Van der Veer (2017) and seek to validate the theoretical frameworks around loss shocks, and their impact on quantity (supply of credit insurance) and premium impact on credit insurance, but by testing data from one of the other “Big 3” credit insurance companies. I hypothesize that an increase in claims is positively correlated to the level of premiums charged, as loss shocks reduce an insurer`s capital, who will compensate for this by charging higher prices, thus a higher premium. Additionally, I hypothesize that an unexpected increase in claims will have a negative impact on the share of exports insured, given that insurers have to allocate capital towards unexpected claims.

Question (3) investigates the impact of loss shocks on premiums and share of exports insured during crisis years and compares this to pre- and post- crisis years. I hypothesize that in crisis years (i.e. 2008 - 2011), the impact of loss shocks should be greater than in non-crisis years (2000 to 2007 and 2012 to 2015), as insurers face increased capital constraints and have limited headroom in absorbing shocks. When comparing results on a pre- and post- crisis basis, I hypothesize that the impact of loss shocks pre-crisis should be higher on the share of exports insured and premiums than post-crisis. Following a crisis, insurance providers have to implement more stringent policy measures and be more cautious in managing their portfolio (Auboin, 2009) .This ultimately means that after the crisis they have been better protected in

(8)

8 the event of a loss shock and would likely be able to better maintain their share of exports insured.

This paper is organized in the following way. Firstly, I will provide a background on trade finance, focusing on private credit insurance and zooming in on its key components including: Claims, Premiums, Underwriting measures and Loss Shock Theory. Following this, a comprehensive literature review analyzing and comparing existing research in this space will be provided. These sections serve as the basis of this paper’s analysis. Furthermore, an overview of the data and methodology including the definitions of key variables and the specifications of the model is discussed. In this section, possible econometric implication will be presented. Following this, I will present the empirical findings together with a thorough discussion of the results. Lastly, the conclusions and implications of my empirical findings are discussed.

3. Theory and Background

This section elaborates on the main theoretical perspectives surrounding the topic of loss shocks in trade credit insurance market. Prior to doing so, it is imperative to define each of the key variables which will appear in my model in the following section:

 Claims: the amount paid out to insurance-holders in an event of default or late payment, and reflects the level of activity in the trade credit insurance market.

 Premium: The risk premium is calculated as a percentage of the total of the customer’s insured sales at a specific periodor of the total insured outstanding amounts at the end of this period

 Loss shocks: an unexpected increase in Claims, in this study represented by Claims Level (Claims / Export Insured) and Claims Ratio (Claims/ Premium).

3.1 Trade Finance

The aim of trade finance is to provide liquidity that facilitates trade and to mitigate the risks involved in trade (Auboin & Meier-Ewert, 2008). The main risks associated with trade finance

(9)

9 relate to the liquidity gap which arises due to a mismatch in timing between the delivery of goods by the seller and the payment for the goods by buyers, as well as potential failure of payment from the buyer all together. Liquidity gaps arise due to the fact that sellers generally incur costs of manufacturing, logistics and delivery prior to receiving the payment for goods/services from the buyer. Thus, if buyers ultimately fail to pay for the goods or services, sellers are left in a vulnerable state due to reduced capital levels and potential risk of financial distress. Failure of payment can be due to buyer-specific reasons (i.e. poor performance, cash flow issues etc.) and is more often than not accompanied with macroeconomic instability, political risk, information asymmetry and moral hazard, all of which increases the need for trade finance insurance providers to step in and neutralize these risks (World Bank, 2009). In addition to this, phenomena such as globalization and international trade have led to increased complexity in trade (i.e. due to increased distances, cultural barriers etc. between buyers and sellers) which further amplifies potential liquidity gaps. Trade finance insurance providers protect sellers from these risks, and thereby provide a critical liquidity function that facilitates trade. They ensure that, in the event of nonpayment, sellers are paid for their goods and services and demand a premium in return. Essentially, the critical role of trade finance providers is to offset the liquidity gap and to ensure that trading takes place.

Research shows that 90% of trade transactions involve a type of trade finance and the overall market for trade credit and insurance has a size of about $10-12 trillion. (Auboin, 2009). The main institution that provide trade finance are banks. The main instruments that bank offers are Letters of Credits (LCs) and documentary collections. Across destinations, the share of sales that are settled with LCs takes values between zero and 90 percent; for DCs, usage lies between zero and 10 percent (Niepmann and Schmidt- Eisenlohr (2013). Apart from banks, two important financial institutions that offer trade insurance are: export credit agencies and private credit insurers.

Export Credit Agency is a financial institution which can be publicly backed by the government or semi government, that provide financial guarantees or insurance to domestic businesses for trading abroad their goods or services. On the other hand, private insurers play similar role as

(10)

10 the Export Credit Agencies, but mostly focused on short term credit insurance. For the latter, private insurers dominate the market in all OECD countries, except Japan and Canada (World Bank, 2009).

The main difference between the above is that private credit insurance covers short term credits with maturity less than 2 years, where ECAs are concerned with long term maturities. Also, ECAs focus on large one-off transactions, whereas private insurers on whole sale turnover. ECAs are involved in international trade, especially with emerging markets where private credit insurers cover both domestic markets and international mainly in OECD countries.

3. 2 Trade Credit Insurance and its mechanism

Trade Credit Insurance follows under the category of trade finance. It is an insurance that covers businesses against the risk that their buyer does not pay their obligation for the goods/services received. Exporters could be risk averse when trading internationally, due to not having enough knowledge on their buyers or uncertainty of the importer country risk and market. International trade developments have emphasized the need of trade insurance for exporters. This insurance policy is offered to domestic and international businesses that want to protect their accounts receivable from loss due to credit risks, protracted default, bankruptcy, solvency and also political risk due to the behaviour of the buyer’s government country (Jones, 2010). The below graph shows the mechanism of trade credit insurance:

(11)

11 Source: Jones, 2010

There are several mechanisms that allow the credit insurer to protect itself against risk. Risk is higher for longer coverage, hence private trade insurer cover risk caused by commercial3 and political4 events for a short period of 90 to 180 days. The insurer has a right to take legal action against the defaulted buyer, which is called the right of subrogation. Furthermore, the insurer sets credit limits and specific terms in the policy on the exporter’s buyer such as by using information on the buyer’s risk. The credit limits set are allowed to be cancelled or changed whenever the insurer decides so. Moreover, the policies do not cover 100 % of the losses, motivating the customer to follow carefully its buyers in case of nonpayment or other risks. The trade insurance policy covers the percentage of the outstanding debt, which ranges between 75% to 95% of the invoice amount, depending on the terms of the contract negotiated (ICISA, 2015). The policy can also have clauses such as: a waiting period where the customer has to do its best to recover the outstanding payments from the buyer and a set threshold that a loss (claim) has to reach, before the insurer pays the claim.

Statistics from Berne Union (2017) indicate that private credit insurance providers have paid out claims of around 56 billion euros worldwide, since the beginning of Global Financial Crisis. In addition to this,

As seen in Figure 1, Claims Paid for short term trade credits have increased from USD 1.5 billion to more than 2.5 billion in the period of 2010 to 2015. Berne Union indicates that the highest volumes of claims paid in 2015 stemmed from defaults in Russia (USD 236 million), Brazil (USD 205 million) and Venezuela (USD 202 million). Since this period, there has been an increased demand for the credit insurance industry, indicating that its role on supporting trade has been of importance even during the crisis. On a bigger perspective, the trade credit insurance fosters economic growth and international trade.

3

Insolvency of the buyer and extended late payment.

4

(12)

12 Figure 1. Claims and Recoveries for short term credits for 2005 to 2015

Source: [Berne Union Statistics. 2008-2013; Berne Union Statistics. 2011-2015; Berne Union Yearbook 2009]

3.3 Claims

Claims paid, which refers to the amount paid out to insurance-holders in an event of default or late payment, and reflects the level of activity in the trade credit insurance market. Even though the aim of credit insurance is not to simply pay claims as they increase, but also to give their support to the customer to mitigate foreseeable risk/losses.

While claims have generally been following an increasing trend, it is important to acknowledge the fact that they are significantly influenced by the health of the broader economy. As evidenced by historical trends, claims paid tend to be influenced by business cycles. For example, the loss ratio (claims paid divided by premiums), increased from an average of 42% in 2004, 2005 and 2006 to 45% in 2007 and 85% in 2008 (see Figure 3) according to data released by ICISA Statistics. Periods of favourable economic and market conditions facilitates a reduction in the claims ratio and periods of instability is conducive to a higher claims ratio. This is more evident when comparing Figure 2 and Figure 3. For example in periods of positive GDP growth (2002 to 2006), claims paid remained flat at approximately EUR 3.0 billion. During the Financial

(13)

13 Crisis (i.e. negative GDP growth in 2008 and 2009), claims increased substantially up to EUR 6.0 billion. These data points emphasize the inherent link between economic activity and claims paid.

Figure 2. Annual GDP Growth ( %)- OECD Member Countries

Source: World Bank website

Figure 2. Premiums, Claims and Claims Ratio

(14)

14

3.4 Premium

The risk premium is calculated as a percentage of the total of the customer’s insured sales at a specific periodor of the total insured outstanding amounts at the end of this period. The credit insurer can charge this in parts based on the declaration of the policy holder (customer) or it can be on a calculated preliminary installment basis with a reconciliation at the end of the policy period. This varies within policies, for example, there are agreements that the premium is fixed for the whole policy period and for some others the premium is calculated on the amount of customer’s credit limits.

The pricing (premium) assigned to a customer is a reflection of the credit risk of its insured portfolio of buyers and the situation of the credit insurance market. This is based on several factors such as: trade sector, the days sales outstanding, buyer countries, customer’s loss history etc.

For newly issued insurance policies, the premium is calculated at the start of the policy on the basis of a prescribed percentage of forecasted turnover for the period of the policy. The policyholders are then obliged to declare actual turnover levels on a monthly, quarterly or annual basis. In cases where actual turnover exceeds forecasted turnover, customers are charged an additional premium percentage above the prescribed rate at the start of the policy. Due to the way this adjustment mechanism works, it must be noted that these adjustments are not immediate, as there could be a significant period of time between the start of the policy and an event in which an adjustment is warranted (i.e., after one year, once it appears that a certain customer’s turnover levels are different from forecasted levels).

3.5 Underwriting Measures

In order to protect themselves from unexpected loss shocks, insurance providers can undertake several underwriting measures that can enable them to smoothen their exposure to downside. For example, insurance providers can set a credit limit for each buyer that the policy of the customer covered. The limit set on the buyers represents the maximum exposure that the

(15)

15 insurer will cover, based on the policy period. In case of a deteriorating risk environment, the insurer can take precaution by withdrawing or reducing the credit limit, without retroactive effect. The insurer will take this action in case the creditworthiness of the buyer has worsened and it cannot be considered an acceptable risk. The private credit insurance contracts are usually “whole turnover” policies, insuring all of a firm’s trade receivables in its buyers, with the choice to exclude risky buyers from the coverage and be able to cancel credit limits (Van der Veer, 2017). Risk is also mitigated through policy underwriting, by including provisions such as Discretionary Limits, that permits the customer to set a credit limit on the buyer without specific review by the credit insurance (in a pre-defined scope).

Another measure insurance providers can take to reduce their exposure to downside risks is reinsurance. Studies show that global insurers transfer the risks associated with approximately half of their business to reinsurers (Jones, 2010). Effectively, reinsurance reduces the need for an insurance provider’s own capital in the event of loss shocks (as these will be covered by the reinsurance party).

The above measures suggest that insurance providers have several methods that can be used in order to deal with loss shocks. Underwriting measures typically revolve around quantity adjustments (i.e. credit limits, discretionary measures etc.), suggesting that loss shocks would likely lead to a proportionately larger adjustment in quantity as opposed to pricing. Policy renewals are not done only once a year, but it can be changed often over the course of the policy period (Jones, 2010). In this way the insurer can constantly monitor the customer and assess the risk. The mechanisms underlying insurer responses to loss shocks are explained in the next section.

3.6 Loss Shock Theory

Academic literature broadly agrees with the fact that loss shocks have a negative impact on the share of exports insured (quantity impact), and premiums (price impact) (Gron, 1994; Winter,

(16)

16 1994; Cummins and Danzon, 1997). Two main theoretical perspectives exist that can potentially explain the mechanisms underlying this.

Loss shocks, such as an unanticipated increase in claims, have a negative impact on the capital levels of an insurer. Thus, in an event of a loss shock, an insurer would be constrained in underwriting insurance products to new clients. In order to compensate for the reduction in capital, the capital constraints theory postulates that insurance providers would have to increase premiums (leading to higher earnings, which would ultimately be used to build up capital levels again) as raising capital from external sources is both costly and time-consuming (Van der Veer, 2017). It must be noted, however, that an increase in premiums following a loss shock, would imply an upward shift of the supply curve, reflecting the higher premiums charged and thus a reduction in the number of buyers of insurance products (as at a higher price point, there will be a lower number of buyers that are willing and able to purchase trade insurance). This dynamic is important as it captures the idea that loss shocks lead to both increases in premiums as well as decreases in the share of exports insured.

Loss shocks may also have an impact on an insurer’s view of losses that could potentially be incurred in the future. In order to account for this, insurers may, following a loss shock, increase premiums on the basis of probability updating (i.e. loss shocks may imply higher probabilities of future losses, which insurers will compensate for by increasing premiums to build a capital buffer) (Froot and O’Connell, 1999; Lai et al., 2000). Thus, similar to capital constraints theory, probability updating also implies that loss shocks lead to capital constraints and an increase in price. A caveat with regard to probability updating is the fact that it may also lead to updated views from the demand side (which would mean that an increase in premiums would not necessarily imply an explicit reduction in quantity, if clients are willing to pay the higher prices).

Overall, both theoretical perspectives support the notion loss shocks lead to higher premiums and a lower share of insured exports.

(17)

17

4. Literature Review

There is a lack of research in the field of credit insurance, and especially in analyzing the effect of a loss shock on the share of export insured and premiums charged. However, authors have explored how a shock in the supply of trade finance has affected trade.

Niepmann and Schmidt- Eisenlohr (2013) investigate the impact of loss shocks on the supply of trade finance and the subsequent effect on US exports. Their study finds economically significant evidence that a reduction in the supply of Letters of Credit has an effect on exports. On average, they find that a loss shock of one standard deviation to a country’s supply of letters of credit decreases export growth by 2 percentage points. The authors explore the LCs in support of U.S. exports, provided by the largest U.S. banks over a period of 15 years. The bank group that has issued the LCs for example to a Spanish importer, has an exposure with a value that corresponds to the total value of the goods exported from U.S. to the importer, this is described as trade finance claims on Spain. During the period that the authors conduct the research there were two peaks in trade finance claims: in 1998 during the Asian Crisis and 2008 during the Global Financial Crisis. The effect of these claims on the supply of trade finance is bigger for exports to small and risky destinations and for industries that depend more on bank guarantees when trading internationally. Moreover, the trade finance claims for U.S. banks have increased dramatically since 2010, mainly due to the fact the some of the European banks went out of this markets allowing U.S. to gain more market share. The two main implications that derive from the findings are: export behaviors of firms can be affected from the pricing and business planning of the main trade finance banks and this affect could be stronger for long-distance trade.

Similar to the above study, Amit and Weinstein (2011) using data from Japan, examine how a shock in the supply of trade finance and credit effects trade. The authors find evidence that this loss shock is transmitted to the exporters and thus it affects their decision on the export. They argue that there is two channels on how these financial shocks can be transmitted to exporters:

(18)

18 financial institutions might have a hard time raising new funds (liquidity problems) and hence rise their rates for trade finance and/or due to losses, these institutions may have inadequate capital levels, leading them to lower the level of coverage provided to exporters. In Japan during the two crises 1990 and 2008, when the interbank borrowing rates resulted to be higher than the government bond ones. The bank of Japan in the financial crisis of 1998 communicated that financial institutions are becoming more cautious in lending due to the constraints in capital. The main regression is composed of two main interest variables: market to book value (bank’s health) and exports. Furthermore, the authors makes use of three types fixed effects: bank fixed effects, year fixed effects and year-industry fixed effects to sweep out macro shocks and to control for supply-and-demand shocks that are common to all exporters in an industry in a certain moment of time. Moreover, the results from their research shows that when banks become unhealthy they lend less; a 1 percent decrease in a bank’s market value is associated with a 0.46 percent decline in trade finance and 0.14 percent decrease in total lending. This shows that the fall in banks health effects supply of trade finance much more than other types of lending. The results indicate that financial (loss) shocks in trade finance play a crucial role in export falls. These findings are important for the development of my research, because it supports the results found that will be introduced in the sections that follow.

The most relevant study to this paper is the research by Van der Veer (2017). Van der Veer (2017) uses bilateral country-level data from 1992 to 2006 from one of the biggest credit insurance companies, to analyze the impact of a loss shock (Claims) on the premium and the share of export insured. The author argues that there has been research on loss shocks in trade finance and its consequences, particularly trade finance provided by banks, but little is known about loss shocks in credit insurance industry. The reason is that credit insurance is a niche market, represented by three private main players, which makes it harder to have access to data. The methodology includes three fixed effects variables: exporter-importer fixed effects, importer-year fixed effects and exporter-year fixed effects. These fixed effects control for all macro shocks that happen and affect insurance underwriting in exporter and importer country. Exporter and Importer year time fixed effects control for supply and demand shocks in both

(19)

19 exporting and destination countries , whereas time invariant country pair fixed effects absorbs common characteristics (i.e. distance) that influence the share of export insured. The results of this research show that a double of claims ratio on average results in a decrease of 11 % of the share of export insure and an increase of 4% of the premiums received by the exporter. The author state that these results are of high importance when it comes to analyzing trade finance constraints during a financial turmoil, since the 3 Big insurers have a dominant presence in the market.

Credit Insurance in Crisis

Auboin and Engermann (2014) uses insured trade credit data from Berne Union members (private and public insurer) for the period 2005 to 2011, quarterly, for about 100 countries. The data is used in a two stage methodology, in the first stage testing the link between financial conditions and trade credit availability and in the second stage, trade credit availability and trade flows. To measure the level of risk of trade credit which is crucial in dictating the supply, the study uses claims on insured trade credit default. From a supply point of view, credit insurer tend to be more cautious in extending credit insurance during risky times, however the authors argue that though this effect is significant it is relatively small. The underlying reason for this, is that credit insurer are inclined to support their customers while at the same time be more carefully in choosing new exposures.

The results show that under the first stage, the volume of insured trade credit available is strongly correlated with economic and financial conditions. The paper points out that the level of liquidity in the economy and GDP are strong determinants on insured trade credit, whereas the risk of trade credit has a small but highly significant effect on the availability of trade credit. Furthermore, in the second stage, it is shown that trade credit strongly determines trade (imports). A 1 percent increase in trade credit provided to a country, results in 0.4 % increase in real imports by that country. The main conclusion of this paper is that trade credit is equally important in a crisis and non-crisis period, this means that the effect that insured trade credit

(20)

20 has on trade stays stable over the cycle. From a policy perspective, there is an emphasis in keeping market incentives high for the supply of credit insurance and encouraging access to trade credit insurance.

5. Empirical strategy and Data

In this section I describe the methodology used in the research and the data used. Firstly I give an overview of the methodology and the specifications of the variables. Following with a description of the data and the reasoning behind the used variables.

5.1 Methodology

The aim of the paper is to identify whether an increase in claims paid (loss shock) by a private trade credit insurer leads to a decrease in the quantity of insurance provided and to an increase in the premium for its supply to the exporter. The analysis is based on data from one of the biggest global credit insurance companies. Following the methodology used by Van der Veer (2017), I use the same regressions and specifications to conduct this research.

The two equation corresponding to my two research questions are the below:

In the above equations, “i” denotes the exporter and “j” the importer (country). As for “t” denotes the time in years.

(21)

21 The dependent variable Share Exports Insured is calculated as a fraction between insured exports over the FOB exports from country i to j. The independent variable ClaimsRatio is calculated as Paid Claims divided by premiums received on insured export from exporter i to importer j. Whereas in the premium equation variable PremiumLevel is calculated as Premiums received divided by the value of insured exports from country i to j. The explanatory variable

ClaimsLevel is normalized by the value of insured exports, thus a fraction between claims paid

and insured export from country i to j. In the Data section I delve deeper into the underlying reason of using this ratios. As for the dummy variables included in the equations, it represent country-pair fixed effects (ij), exporter- year fixed effects ( it) and importer year fixed effect (jt), which cover the unobserved influences on loss shocks , share of exports insured and premium level.

Based on the above equations, the two main parameters are B1 and B2 that indicate the average effect of claimsratio on share export insured and the average effect of claimsratio on

premiumlevel, expressed as elasticities.

The methodology used is Ordinary Least Squares and fixed effects. In order to tackle heteroskedasticity I use a robust covariance estimator, clustered by the country-pair (Exporter-Importer Fixed Effect). The inclusion of fixed effects are essential to absorb any factors that stems from insurance underwriting or other influences rather than claims. The time-varying exporter-importer fixed effects together with the country-pair fixed effects, are commonly known from the gravity model that is used for predicting international trade level. This model is used from study of Van der Veer (2017). This model uses a panel data of bilateral trade agreements between exporter and importer countries. Recent econometric evaluations of this model with panel data, as for example Egger (2000) using the Hausman Test , have concluded a rejection of random effects gravity model compared to a fixed effect ( country time variant and country-pair) gravity model (Baier and Bergstrand, 2007).

(22)

22 Though the equations in this study are not based on the gravity model theory, the specifications are comparable (bilateral time series data between exporter-importer). As such the fixed effects are of high importance in the context of this methodology. Two types of fixed effects are implemented into the model being used: country-based fixed effects (i.e. country pairs) and time-based fixed effects (for both importers and exporters).

These fixed effects are used to cover unobserved heterogeneity that will influence my results. For example, the exporter and importer-year FEs will eliminate the potential time-varying country-specific shocks in supply and demand within respective insurance markets (i.e. these supply / demand shocks may overshadow the impact of loss shocks which is the main variable of interest). The time-based fixed effects mainly account for impact of business cycles on insurance underwriting in the countries of origin of both exporters and importers. Claims tend to increase in periods of economic recession (i.e. business cycle downturns) and decrease in periods of economic booms (i.e. business cycle upturns) and hence these exporter- and importer- year fixed effects ensure that business cycle changes are not picked up. The country pair fixed effects control for any characteristics (time – invariant) that are common to the exporter and importer (such as distance) and could bias the results.

5.1.1 Extensions to the benchmark model

Firstly, in order to test the duration of a claim effect on the share of exports insured and premium, I add lags of claims to both of equation (1) and (2). The equations have the same specification as the benchmark model.

(23)

23 Secondly to examine the crisis influence, I take two approach. I split the sample in two manners: pre- and post- crisis (first test), and pre-, during, and post- crisis (second test) in order to see if there are meaningful differences on premiums and shares of exports insured. The first split is based on the structural break test that shows that there is a structural break in 2008 when looking at claims. The second split done on the data is based on the paper by Dedi and Yavas (2017), I explain further on the crisis section their analysis.

5.2 Data

The data of this research is retrieved from the database of one of the three global private credit insurance, for confidentiality matters it cannot be named. This data base is not the same one used on the study from Van der Veer (2017).

The panel data is composed of bilateral data for a certain exporter and importer in a specific year, including the variables of interest which are the aggregate Premium received, Export Insured and the Claims Paid. The data set includes also variable specific to exporter or importer such as GDP, population, region etc., which are not of main interest in the methodology. Based on the variables of interest, the database is restricted to 8,462 observations, after the cleaning of the data. The cleaning of the data involves, removing all the observations that do not have insured export coverage, all the observations that do not have Claims and Premium data and the observations that have a share of export insured higher than 1.

The data is of the 2000-2015, which includes also the Global Financial Crisis. I have chosen to take exporter OECD countries due to the fact that there is more credible and available data,

(24)

24 since private credit insurance is more present in this market. As for the importers, the list amounts to 189 countries worldwide. See Appendix 1 for the list of exporters and importers and the frequency of the observations.

Insured export is referred to as the export that has risk coverage from one of the biggest private credit insurance globally, as mentioned above. The insured export is different for each export-importer pair, and has a mean of EUR 173 million and a maximum of EUR 8.5 billion. The share of export insured represents the cover that the insurance company has given to a certain exporter for a specific importer. The share of export insured is calculated as total f.o.b 5 Export provided by IMF 2017, divided by the goods export insured. This share has a mean of 6 %, but the maximum goes to 99%. I deal with these extreme ends in the sensitivity analysis and I explain below the flaws of the share of export insured data.

Claims Level is a known measure to check the loss of the credit insurance, by the private insurers this is also known as the “loss ratio”. The combined ratio includes expense ratio plus claim ratio divided by the revenues of the insurer. A study done by AU Group (2016) points to the fact the theoretical break-even point of insurers in the context of loss ratio is 70 % (i.e. if the loss ratio exceeds 70 % insurers would consider that their operations are making a loss). Statistics from ICISA during the years, show that the expense ratios for the main insurers vary from 20 % to 40 %. The claim ratio during 2003 to 2007, was around 45%, but jumped up to 85 % of ICISA members and for some above 100% (Zurich American Insurance Company, 2014). There is a huge skewness on the claims ratio in the data and in order to tackle this I take logs of the variables. Moreover I test in the sensitivity analysis that the outliers do not drive the results. The median Claims Ratio is 58 %, which is close to the average market Claim Ratio of 52 % reported by ICISA (2016) for the period 2001 to 2014.

In the premium equation the two variables used are premiums level and claim level. Premiums Level is calculated as premium received divided by insured export, the mean of this variable is

5

(25)

25 0.23 %, and ICISA (2016) reports an average premium level of 0.31 percent for the trade insurance market for the period 2005 to 2014. The claims level is calculated as claims paid divided by insured export and ICISA (2016) reports an average claims level of 0.16 %. From the data of this research, the average premium level 0.14 percent and claim level is 0.11 percent.

Worth mentioning is that the database is not a creation of the author, but the cleaning and adding of some of the variables of interest is part of the work done for this research. See Appendix 2 for summary statistics. Summary Statistics, include the creation of new variables such as Share of Export Insured, Premium Level, Claim Level, Claim Ratio and the lagged value of the Claims Ratio and Claims Level. All data with Share of Export Insured have been disregarded.

There are some specific modifications that there have been done in the dataset, based on plausible explanations. Firstly, the observations that did not include any data on exports and insurance coverage between two countries have been disregarded, because it is not relevant to the study. Thus, the dataset only includes bilateral observations between exporter-importer that for certain year had an insurance. Moreover, all the observations that have a share of export insured bigger or equal to 1, have been removed from the database. I explain the reasoning behind this on the section that follows.

5.3 Potential Econometric Issues

1. Endogeneity

Endogeneity is an econometric issue that can occur when the explanatory variable is correlated with the error term. It can also occur due to measurement errors, inverse causality or omitted variables.

(26)

26 Based on Van der Veer (2017) and Baier and Bergstrand (2007) in the equations of this paper I include three pair of fixed effects that account for all time-invariant country pair characteristics, and all time-varying exporter-importer. The rich set of dummy variables controls for any omitted variable bias that would result from time-varying country pair characteristics. Whereas time- invariant fixed effects control for any omitted variable bias that could stem from characteristics that are common to both the exporter and importer.

1. Share of insured export higher than 1

The database includes transactions between exporter and importer that have a share of insured export higher than one. These observations compose around 2% of the database. This measurement error can stem from an overestimated numerator, insured exports. If the insured export is not a correct representation of the export insured from country I to j, this could lead to bias in the numbers. Based on a research of the private insurer, financial transactions seem to be a big representative of insured export, but at the same time also have the least number of transactions. Financial transactions and transaction that are not goods export could be all possible sources of measurement errors. This leads to an overestimation of the true insured exports. Furthermore, the billing country is not always the country of exports and in some cases the subsidiaries are clustered and declared as sales to one country. In both cases, insured exports between two countries could be overestimated or underestimated, contaminating the data.

Looking at the denominator, the export value could be biased if the value also includes service exports. Based on IMF (2018), exports are measured as FOB exports and cover only goods exports. The inclusion of service exports, will bring the share up. Countries such as Ireland, Denmark that have a high share, have a bigger proportion of services included in the exports.

The high shares make a very small part of the samples, and to check whether these high shares could bias the results, I conduct a test in the robustness check.

(27)

27 2. Robustness Checks

In order to check for the robustness of my results, in the Sensitivity analysis part, I have estimated the model again by making several changes to the sample. These changes include dropping observations based on geographical region, GDP, population sizes of the importer. Due to the fact that the market for credit insurance if mostly focused within Europe, I check the robustness of the results by testing the effects only when there is a European importer and exporter.

Furthermore, as mentioned above there could be bias due to share of export insure higher than one and also there is skewness in the Claims ratio due to measurement errors of the insured export. To test whether my results are biased due to these outliers I drop observations of Claims Ratio that are higher than the median value and observations of the Share of Export Insured that are higher than the ratio 0.3 (this number is taken as a benchmark from the business). These results are discussed and presented in Section 7 of this paper.

6. Benchmark Results

6.1 Main Regressions

The below estimates (Table 1 and 2) are in line with theory expectations, thus an increase in the claims ratio, would lead to a decrease in the share of export insured and increase in the premium charged by one of the biggest private credit insurance which is not the same credit insurance as the one of the study from Van der Veer (2017). The results are in line with Van der Veer (2017) research and also with the loss shock theory prediction. As stated in Section 3, a loss shock reduces the available capital of the insurer, forcing them to cover their losses through either internal capital (retained earnings) or external capital (i.e. capital markets). The insurer, as a result, will reduce the provision of coverage to customers (quantity effect) and increase premiums (price effect) in order to compensate for their losses. The above results

(28)

28 (Table 1 and 2) together with the results of Van der Veer (2017) give a clear picture of the market, since the data comes from two of the biggest credit insurance in the market. Van der Veer (2017) states that he can not conclude that the results are applicable to the whole industry. Van der Veer (2017) finds that a doubling of the claims ratio on insured exports from country i to country j leads to an average decline in the share of export insured by 11% and an increase in premium by 4%. The results presented are comparable to the results are Van der Veer (2017), even though the insurer is different and the period of time is from 2000 to 2015. The similarity to Van der Veer’s results possibly reflects the oligopolistic nature of the market (competition between the 3 big players considered “cut throat”), and the overlap between the way the big 3 players operate (i.e. comparable business characteristics and relatively large market shares relative to the broader industry).

Table 1. The effect of an increase in Claims on the Share of Export Insured

(1) (2) (3) (4) (5) (6) (7) (8) VARIABLES lnShare Export Insured lnShare Export Insured lnShare Export Insured lnShare Export Insured lnShare Export Insured lnShare Export Insured lnShare Export Insured lnShare Export Insured lnClaimsRatio -0.314*** -0.189*** -0.351*** -0.252*** -0.176*** -0.246*** -0.118*** -0.110*** (0.00694) (0.0152) (0.0287) (0.0255) (0.0192) (0.0317) (0.0170) (0.0191) Constant 0.950*** (0.0154) Observations 8,462 8,433 8,049 8,056 8,019 7,612 8,014 7,568 R-squared 0.195 0.677 0.352 0.714 0.784 0.785 0.869 0.904 Exporter_Year_FE No Yes No No Yes No Yes Yes Importer_Year_FE No No Yes No Yes Yes No Yes Exporter_Importer_FE No No No Yes No Yes Yes Yes

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

I start my model by a simple Ordinary Least Squares regression, without any fixed effects. From the above results (Table 1), it shows that a double average increase in Claims Ratio leads to a decrease of 31.4% of the Share of Export Insured. The estimate is significant at 1% level, but Rsquared is very low. This is a naive way to look at it because of possible omitted variable bias

(29)

29 which indicates that the differences within the exporter, importer and year should be investigated. Thus, I implement exporter country and year fixed effects into my model (regression (2)). The effect of the lnclaimsratio on lnsharedexport dramatically drops in absolute terms (smaller in negative terms). The results indicate that the average effect of lnclaimsratio on lnsharedexport significantly increases (but remains negative) and secondly the goodness of fit coefficient increases form approximately 20% to 70%. It is important to note that the number of observations slightly decreases which points toward a well-balanced dataset. The inclusion of FEs lowers the number of observations because there is no uniformity in the pairs, some countries may have more pairs than others. A well-balanced dataset shows that for each year and exporter country there is a highly populated dataset. This observation allows us to draw reliable conclusions.

To make it easier in analyzing, I assume that the claim ratio doubles and moreover since the coefficients are in natural logarithm, the coefficients of interest have to be seen as elasticities.

In regression (3), when controlling only for importer-year fixed effect, a doubling of the Claim Ratio, decreases the share of export insured by 35%, this decrease is much more pronounced when this fixed effect is used compared to the exporter year fixed effect used in equation (2) and far from the result of the complete model (8) . This means that the country of the exporter plays a significant role in the relationship of claims ratio and shared of export insured. A similar result is observed in equation (4), where only country-pair fixed effects are included. The coefficient estimate remains negative, even though it changes.

From equation (5) to (7) I alternate between the three fixed effects, by using two of them simultaneously in different combinations, the effect is still negative but however as before when not controlling for export year fixed effect, the average effect of the Claims Ratio on Share of export insured, becomes more negative. Again, this shows that investigating the impact of the exporter countries in this relationship is of great importance. Furthermore,

(30)

30 equation (5) to (7) give a strong indication of the importance of exporter-year effect and importer-year effect.

In equation (8), when controlling for exporter-year, importer-year and country pair fixed effects the average effect is 11%.

From the above analysis of the results, it is evident that results are indeed driven by supply and demand shocks in exporting and importing countries, which have to be controlled for through exporter-year and country pair fixed effects. The presence of exporter-year fixed effects (column 2 of Table 2) evidently reduces the coefficient most significantly when comparing it to the base model (column 1), implying that loss shocks have a less negative impact on share of exports insured when controlling for time-varying supply and demand shocks in exporting countries. As also discussed by Van der Veer (2017), the reason this is observed could be related to relative demand; i.e. that the rise of claims might induce exporters to lower the coverage they have from the private credit insurance due to possible higher prices. This result suggests the importance of demand-side influences on the quantity equation, thus making it imperative that we control for exporter-year fixed effects. The demand-side effects are not covered in detail in this study but would be an interesting avenue for future research given the important implications this has on the share of exports insured.

Using the same strategy I estimate the average effect of a doubling of ClaimLevel on the Premium (Table 2). Comparing the first equation where no fixed effects are used to the last ones, the coefficient goes down by 5%, indicating the presence of omitted variable bias if not controlling for country-year and country-pair fixed effects. From equation (8) it can be concluded that a double of claimslevel increase the premiumlevel with 3.6 %. This coefficient is significant at all significance levels

(31)

31

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Importer-fixed effects have a much larger impact on the premium equation relative to exporter-fixed effects. As discussed in Section 3.4, premiums are charged on the basis of two main pillars: i) level of credit risk as a function of the portfolio of importers with which the exporter engages with, and ii) the exporter’s own performance over the period of the policy (Jones, 2010). Thus, premiums tend to increase if i) deterioration in the quality of the portfolio of importers and/or ii) there is a mismatch between forecasted performance and actual performance of the exporter itself (Jones, 2010). Evidently, these results indicate that controlling for importer-fixed effects (i.e. controlling for a sudden deterioration in credit risk of an exporter’s portfolio of buyers by implementing importer-year FEs) has a profound impact in

reducing the effect of loss shocks on premiums charged (i.e. coefficient decreases from 9.0% in

(1) to 5.1% in (3)). When implementing exporter-fixed effects, the impact is less profound (coefficient decreases from 9.0% in (1) to 7.3% in (2)). This ultimately suggests that from a premium perspective, the performance and quality of the portfolio of buyers that an exporter engages with is more important than the exporter’s own absolute performance.

The result are in line with the theory of policy underwriting, that states that the insurer will increase the premium as a countermeasure for an increase in claim ratio/level. Other Table 2. The effect of an increase in Claims on the Premiums

(1) (2) (3) (4) (5) (6) (7) (8) VARIABLES LnPremium Level LnPremium Level LnPremium Level LnPremium Level LnPremium Level LnPremium Level LnPremium Level LnPremium Level LnClaimsLevel 0.0897*** 0.0725*** 0.0510*** 0.0569*** 0.0400*** 0.0452*** 0.0310*** 0.0364*** (0.00377) (0.00457) (0.00673) (0.00432) (0.00545) (0.00683) (0.00568) (0.00712) Constant -5.827*** (0.0272) Observations 8,462 8,433 8,049 8,056 8,019 7,612 8,014 7,568 R- Rsquared 0.063 0.440 0.318 0.477 0.553 0.555 0.606 0.639 Exporter_Year_FE No Yes No No Yes No Yes Yes Importer_Year_FE No No Yes No Yes Yes No Yes Exporter_Importe

r_FE

(32)

32 countermeasures include; policy cancellations and/or possibly restructuring of loss-making policies and limit restrictions. Moreover, based on probability updating theory (Lai et al., 2000), a rise in claims will lead the insurer to increase premiums due to the fear of future losses. Even though theory predicts that a loss shock will lead to an increase in premium, from the above result it can be seen that this elasticity is relatively low (3.6%, in the fully-loaded model in column 8), though statistically significant. The insurer can only renegotiate the premium after one year and most of the contracts premiums are fixed with a duration of two to three years.

There could be several reasons why we observe this phenomenon. Firstly, aggressive competition in the market and the fact that the insurer cannot re-negotiate the premium of a policy directly after the loss shock, but only after the period which the contract expires (Van der Veer, 2015) leads to limited headroom in increasing premiums. Secondly, the re-negotiation of the premium can cause the customer to terminate their policy if the premium asked will be too high, thus the insurer would not increase the premium by a significantly high percentage. Also, insurers tend to use non-premium based measures as well in order to deal with loss shocks (i.e. withdrawing credit limits, implementing additional covenants / restrictions onto exporters etc.).

6.2 Lagged Values

In this section I estimate the average effect of claims on the premium and share of credit insurance, and in the specification I include lags of claims to examine the duration of the impact of a loss shock (Table 3 and Table 4). The negative average effect of an increase in the

claimsratio lasts for about four years. Whereas, the positive average effect of the increase in claimslevel on the premiumslevel dies within the second year from the loss shock. After three

years, the average effect of a double claimsratio on the share of insured export, is only 1.5% and when including a fourth lag it becomes 1% and insignificant. This effect on the premium for the second year is insignificant and down to 0.127%.

(33)

33 Table 3. The effect of lagged Claims on Share of

Export Insured (1) (2) (3) (4) (5) VARIABLES LnShareExpor tInsured LnShareExport Insured LnShareExport Insured LnShareExport Insured LnShareExport Insured lnClaimsRatio -0.110*** -0.0951*** -0.0722*** -0.0706*** -0.0702*** (0.0191) (0.0119) (0.0136) (0.0159) (0.0181) lnClaimsRatio_1 -0.0456*** -0.0444*** -0.0359*** -0.0375*** (0.00661) (0.00776) (0.00925) (0.0105) lnClaimsRatio_2 -0.0249*** -0.0295*** -0.0238** (0.00620) (0.00877) (0.0100) lnClaimsRatio_3 -0.0151** -0.0106 (0.00747) (0.00940) lnClaimsRatio_4 -0.0144** (0.00641) Observations 7,568 4,901 3,705 2,998 2,512 R-squared 0.904 0.925 0.935 0.933 0.932

Exporter_Year_FE Yes Yes Yes Yes Yes

Importer_Year_FE Yes Yes Yes Yes Yes

Exporter_Importer_FE Yes Yes Yes Yes Yes Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 4. The effect of lagged Claims on Premiums

(1) (2) (3) (4)

VARIABLES LnPremiumLevel LnPremiumLevel LnPremiumLevel LnPremiumLevel

lnClaimsLevel 0.0364*** 0.0209*** 0.0213*** 0.0209*** (0.00712) (0.00393) (0.00434) (0.00484) lnClaimsLevel_1 0.00971*** 0.0166*** 0.0154*** (0.00367) (0.00403) (0.00457) lnClaimsLevel_2 0.00127 0.00914* (0.00445) (0.00499) lnClaimsLevel_3 -0.00224 (0.00436) Observations 7,568 4,904 3,706 2,998 R-squared 0.639 0.884 0.895 0.902

Exporter_Year_FE Yes Yes Yes Yes

Importer_Year_FE Yes Yes Yes Yes

Exporter_Importer_FE Yes Yes Yes Yes

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(34)

34 As described in the theory section, insurers continuously review customer policies in order to adjust premiums on a quarterly or yearly basis. This adjustment, however, does not take place immediately (there is a lag due to the review and renegotiation process). The same reasoning is applied to the share of export insured. After a certain period following the loss shock (i.e. one quarter of one year), the insurer may decide whether coverage should continue to be provided for a certain exporter. However, it must be noted that even though this is true in most cases, trade credit insurance providers retain the right to withdraw or terminate the provision of insurance at their own discretion if there are significant changes to the economic, financial or political situation. For example, during the global financial crisis, trade credit insurance providers were able to immediately cut credit limits to protect themselves and their customers.

These results are different from Van der Veer (2017), who finds that the average claims effect dies away after a year for share of exports insured and the effect on the premium dies within the first year after the increase in claims. A possible explanation behind the different results could be the fact that Van der Veer's (2017) data is for the period 1992-2006 which does not include an exceptionally risky environment, whereas the data in this study covers the period 2000-2015, which includes the global financial crisis. The fear of the persistence of the crisis together with the inability to immediately recover losses may explain why the effect of claims has a longer persistence in premium and share of export insured. In addition to this, the impact of a large loss shock during the 1992 - 2006 period may have a different impact than a sizable loss shock in the 2000 - 2015 period. The duration of the impact of the loss shock thus may be linked to such factors (i.e. size of loss shock, time period covered), which differ across the studies. The section that follows provides further analysis on the crisis.

6.3 Further analysis Global Financial Crisis (2008-2009)

The data of this research includes observations from the Global Financial Crisis 2008-2009. The model used in this part does not include lags because it reduces the number of observations, which is already reduced due to the separation of data into two samples. To test of a structural

(35)

35 break in my data, I firstly generate a unique ID for all possible combination for exporter, importer and year and then conduct the test. The result of the test show that the hypothesis that there is a structural break in 2008 cannot be rejected (Appendix 3). Structural break tests shows whether a times series changes abruptly at a point in time. This test showed that data in 2008, which corresponds to the Global Financial Crisis, had an unexpected shift.

Figure 2 plots the mean premiums, claims paid and claims ratio of the all exporter countries from my sample over the entire sample period (2000 - 2015). The results of the structural break test can be supported by the data illustrated on this chart. Clearly, the claims paid and claims ratio experiences a surge in the periods 2000 - 2002 and 2008 - 2009, which were both periods reflecting a deteriorating risk environment in the economy (Dot Com Crisis and Global Financial Crisis respectively).

During the Global Financial Crisis, global trade volumes experienced a sharp decline on the back of all-time low investor confidence and uncertainty. While a plethora of companies were experiencing defaults on their obligations (which was inevitably a major influence in leading to a spike in claims paid as well as the claims ratio, which almost reached 200% in 2009), the resultant reduction in share of exports insured was not so drastic. This is possibly due to the influx of exporters that were still willing to trade and were now triggered to buy credit insurance to protect themselves during this period.

As can be seen in Figure 4 (my dataset) and Figure 5 ( aggregate market data from ICISA, which is comparable), following the Dot Com Bubble crisis, there is a spike in the claims ratio which progressively goes down as market conditions improve in the run up to the financial crisis. 2008 and 2009 see another large spike in the claims ratio before falling to levels below 50% from 2010 onwards. The period 2010 to 2015 sees a progressive decrease in premiums (reflecting higher competition) and a moderate increase in claims, which is also comparable to the market.

(36)

36 Figure 4. Premiums, Claims and Claims Ratio from the trade credit insurer of this research

paper

Figure 5. Premiums, Claims and Claims Ratio from the global trade credit insurance market

Source: ICISA 2001-2015

Furthermore I plot the means of Claims Paid, Insured Export and Premiums for 2000 to 2015 (see Figure 4), which all present results that are generally in line with expectations. There is a sharp increase from 2008 until 2009 of Claims paid and a significant drop in Insured Export

(37)

37 which results in a subsequent significant drop in Aggregate Premiums (i.e. total EUR amount of premiums received from all of the outstanding policies). The main reason for the decline in premiums is quantity-driven. During the financial crisis, insurance providers were able to cancelled credit limits and withdraw their positions due to the huge losses they were experiencing (Van der Veer, 2015).

There are two approaches I look at for analyzing the effect before and after the crisis. Firstly based on the structural break result (Apppendix 3), that shows that there is a break in 2008, I divide my data into pre-crisis (2000 to 2007) and post-crisis (2008 to 2015), the results are shown in Table 5. Secondly I break my data into three parts before crisis (2004 to 2007), during crisis (2008 to 2011), and after the crisis (2012 to 2015). Using a crisis dummy due to the time fixed effects causes multicollinarity thus this approach cannot be taken.

The results in Table 5 compare the effect of a loss shock on premiumlevel and share of export

insured, before and after the crisis. Comparing the regression of claimsratio on share of export

insured for before and after crisis, the average effect is higher in absolute terms for the period 2008 to 2015, with a lower standard error and significant at all significance levels. These results control for exporter, importer and country-pair fixed effects (i.e. eliminating the impact of supply & demand shocks as well as business cycle influences). These results are also supported by the figures published by the Berne Union and ICC in 2010. These indicate that the last quarter of 2008 and the first half of 2009, the coverage offered by private insurers declined by 16 percentage points more than by public agencies. It must be noted that while this decline is mostly related to a decline in supply, other factors such a price and demand may have also contributed to this (ICC, 2016).

As for the effect of claimslevel on premiumlevel, the average effect is higher for the period 2000 to 2007 then for after the crisis (a difference of 2.3%). The reason for this could be related to the measures that the insurer took after the crisis. After the crisis, private insurers cut credit limit and put more emphasizes on choosing their customers, in order to avoid risk rather than

(38)

38 immediately increasing prices by a high percentage. Moreover, competition during the years after the crisis has been very aggressive. As Robert Nijhout, executive director of International Credit Insurance & Surety Association (ICISA) has stated in 2015: “As is the case in other sectors, competition often leads to lower rates, in spite of a risk environment that has stayed the same and in some cases has increased”.

Table 5. The effect of Claims before and after the Crisis

2000-2007 2000-2007 2008-2015 2008-2015

VARIABLES lnShare

Export Insured

lnPremium Level LnShare Export Insured LnPremium Level lnClaimsRatio -0.0829** -0.100*** (0.0321) (0.0172) lnClaimsLevel 0.0417*** 0.0245*** (0.0147) (0.00430) Observations 3,726 3,726 3,536 3,536 R-squared 0.931 0.604 0.912 0.867

Exporter_Year_FE Yes Yes Yes Yes

Importer_Year_FE Yes Yes Yes Yes

Exporter_Importer_FE Yes Yes Yes Yes

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

The analysis presented in Table 6 gives a more thorough picture of the situation before, during and after the crisis. The rationale behind splitting the sample in such a manner is due to the fact that business and investor alike change their behaviour in response to changes in business cycles. Dedi and Yavas (2017) study the equity returns and volatilities before, during and after the crisis. To test their hypothesis the authors separate the dataset in three periods: 5 years before the crisis, a period during the crisis and five year after the crisis. Although the results show significant co-movements of returns in the three periods studied, important differences are noted before and after crisis, in both returns, volatility and diversification opportunities.

Referenties

GERELATEERDE DOCUMENTEN

With intertrmporal contracting, a bank need not aake zero expected profit on a given borrower ín each time period. It can tax the borrower i n one period ar:d subsidize it in the

Wat als eerste opvalt aan het voorgaande is de tweestrijd tussen de twee klassieke pa- radigma’s. Of eigenlijk het ontbreken van deze strijd. Conflict tussen de beide

It is my honour to welcome each and every one of you to this colloquium on Youth Day 16 June 2017, which is organised by the Commission of Faith and Order of the World Council

In comparison, column (3) shows that one percent increase in local access to finance increases bank credit that are given to formal firms by 1.28 percent of their sales value

With that regard, in this study the Indians minority and one of the tribe, Chaggas 2 in Tanzania which is prominent in business activities like Indians are compared to

It is found that when a supplier holds a high level of supplier power, trade credit terms are less attractive compared to a situation in which a supplier holds a lower level of

The product of risk premiums given in Table 23 with the number of persons insured in every cell can be used as row factors in the Bornhuetter-Ferguson method in section 4.3 for

This implies that younger firms are more selfish since they have higher levels of accounts payable to total assets by receiving more trade credit than older