The Relationship between Activities of the Insurance Industry and Economic Growth: The Case of G-20 Economies

Abstract

The aim of this study is to examine the impact of developments in the insurance sector on economic growth. This study examines an annual dataset spanning 1992 to 2022 from ten countries within the G-20 economies, for which data is reliably accessible. The analysis employs panel cointegration and panel causality tests to explore the underlying relationships. The results show that the activities of the overall insurance sector have a positive effect on economic development, while non-life insurance activities, as a sub-component of the insurance sector, also demonstrate a similar impact. Furthermore, findings indicate the existence of a bidirectional causality relationship between economic development and non-life insurance premiums. This study provides various recommendations to policymakers in the process of developing policies on the growth of national economies.

1. Introduction

Significant transformations have been observed in the global insurance sector in recent years. Influenced by a combination of technological advancements, economic factors, and changing customer preferences, the industry has experienced a robust growth trend. While this trend slowed during the COVID-19 pandemic, it has since accelerated alongside the emergence of insurtech, which refers to the application of information technology in the insurance sector [1]. According to OECD data, the global general insurance market reached a size of approximately $5.94 trillion in 2022 and is projected to reach around $6.80 trillion in 2023. The increasing awareness of the importance of insurance, driven by rising income levels, the growing significance of emerging markets particularly in Asia and Africa and the focus of insurers on developing specialized products and distribution strategies to meet the unique needs of these markets, play a crucial role in the expansion of the market [2].

The insurance sector, which assists individuals, businesses, and governments in managing various risks and uncertainties and serves as a critical component of the financial system, plays a vital role in supporting economic growth. By providing a range of services and products that help mitigate risks, protect individual and commercial assets, and facilitate financial stability, the insurance sector contributes to economic growth through various channels [3,4]. The risk management services offered by the insurance sector can reduce uncertainty and volatility associated with economic activities, thereby promoting investments [5]. Furthermore, whether stemming from individual incidents such as accidents or illnesses, or larger-scale disasters like natural calamities or financial crises, the insurance sector enables businesses and individuals to focus on their operations efficiently by mitigating the impact of unexpected losses, free from the fear of financial ruin [6]. The insurance sector also facilitates the growth of various industries by providing necessary risk coverage and financial security through products such as property, liability, and life insurance [5]. Additionally, the sector contributes to economic growth through direct and indirect employment effects. In this context, insurance companies operating in the market create employment opportunities that generate income and economic activity both within the sector and in supporting industries [7].

In conclusion, the insurance sector is a vital component of the financial system that contributes to economic growth through its role as a financial intermediary, its ability to mitigate risks, its support for other economic sectors, and its employment-generating effects [4]. However, the growth of the insurance sector is closely linked to the overall macroeconomic conditions of the country. As the economy grows, the demand for insurance products and services tends to increase as individuals and businesses seek to protect their growing assets and incomes [8].

Despite the economic and geopolitical turbulence experienced worldwide in recent years, the observed growth trend in the insurance sector suggests that the role of insurance activities in the economic development of countries should be re-evaluated. In this context, a group of countries has been selected from the G-20 economies, which possess significant economic scale and for which reliable data is accessible. This group includes Australia, France, Germany, Italy, South Korea, Mexico, Japan, Turkey, the United Kingdom, and the United States. The study aims to examine the relationship between growth in the insurance sector and economic growth. As previously mentioned, it is believed that there exists a complex relationship between growth in the insurance sector and economic growth. The presence of various studies in the literature that yield differing results supports this notion.

The remainder of this study is organized as follows: Section 2 provides an overview of previous studies addressing the relationship between economic growth and the insurance sector. Section 3 explains the model and the data utilized. Section 4 presents the findings obtained from the study, while the Section 5 offers a summary of the study’s results and policy recommendations.

2. Literature Survey

This section compiles the existing literature examining the relationship between the insurance sector and economic growth. An examination of the existing literature reveals various studies that address whether the growth and development of the insurance sector significantly impact a country’s economic performance. These studies can generally be categorized into three main headings. The first corresponds to supply-leading theory, which examines the unidirectional effect of insurance activities on economic growth; the second corresponds to demand-following theory, which addresses the inverse relationship; and the third encompasses studies that test both supply-leading and demand-following theories while considering bidirectional causality.

Regarding supply-leading theory, ref. [9] analyzed panel data from 12 industrialized countries for the period 1970–1981, investigating the relationship between property-liability insurance premiums and income, and provided evidence of a relationship between economic growth and non-life insurance. Ref. [10] examined the impact of insurance on capital and output growth, concluding that insurance premiums have a significant effect on economic growth. In another study, ref. [11] investigated the effects of both insurance investments and premiums on GDP growth in Europe, conducting a panel data analysis on data from 29 European countries for the period 1992–2005. The study found that life insurance had a positive effect on GDP growth for 15 EU countries. Ref. [12] conducted a panel data analysis on 55 countries from 1976 to 2004 to explore the relationship between the insurance sector and economic growth. Arena considered insurance premiums as indicators of the sector’s growth and concluded that both life and non-life insurance types positively influence economic growth. Ref. [13] tested the relationship between insurance sector activities and economic growth using ARDL and FMOLS methods, revealing that non-life insurance penetration in Ghana has a stronger positive relationship with economic growth compared to life insurance penetration. They also noted that non-life insurance activities contribute more to long-term economic growth. Ref. [14] investigated the causality relationship between insurance sector activities, banking sector activities, and economic growth in G-20 countries from 1980 to 2014, providing evidence of a causal relationship between insurance sector activities and economic growth. Ref. [15] Similarly, identified a causality relationship between insurance sector activities and economic growth in China, concluding that homeowners’ insurance is more effective than life insurance in promoting growth. Additionally, ref. [16] found evidence supporting the connection between insurance sector activities and economic activities in their study conducted across 90 countries. Based on their findings, they suggested that the development of the insurance sector could contribute to economic growth, although this relationship may vary according to different income levels and types of insurance.

Studies highlighting the existence of an inverse relationship corresponding to demand-following theory include the work of [17]. This study utilized a dataset encompassing 45 developing countries, focusing on the correlation between life insurance demand and per capita GDP. It was found that as a country’s per capita GDP increases, so does the demand for life insurance. Similarly, ref. [18] identified a positive relationship between non-life insurance demand and per capita GDP, supporting the notion that economic growth influences insurance consumption patterns. Ref. [19] examined the relationship between economic growth and the insurance sector in their study involving nine OECD countries. Utilizing annual data from 1961 to 1996, they tested the causality relationship between real GDP and real insurance premiums. Their findings indicated that the magnitude and direction of the causality relationship vary by country. In some countries, the insurance sector was found to be a cause of economic growth, while in others, economic growth was identified as a cause of the development of the insurance sector. Ref. [20] conducted a study focusing on the determinants of life insurance consumption in 63 countries from 1980 to 1996, shedding light on the factors influencing international demand for life insurance products. Their findings revealed insights into the relationship between insurance demand and economic indicators. In another study, ref. [21] investigated the causality between developments in the insurance sector and economic growth in 19 developing Asian economies from 2007 to 2017. Their findings concluded that better economic growth, stable exchange rates, positive price changes, and age dependencies contribute to the growth of the insurance sector in developing economies. Ref. [22] found that economic growth positively affects life insurance consumption, particularly in middle-income economies, supporting demand-following hypotheses.

Many existing studies in the literature focus either on the impact of economic growth on insurance activities or, conversely, on the effects of insurance activities on economic growth. However, there are also studies that provide evidence for the bidirectional relationship between these two concepts. For instance, ref. [23] examined the causality relationship between life insurance and GDP in the United Kingdom using nine different types of life insurance premiums. They found evidence of long-term causality from insurance premiums to GDP in six cases. In one case, they identified causality from GDP to insurance premiums, while in two other cases, a bidirectional causality relationship was established. Ref. [24] investigated the short-term causality between the life and non-life insurance sectors and economic growth in India, revealing a bidirectional causality between the development of the life insurance sector and economic growth. Similarly, ref. [25] included 41 countries with three different income levels in their study covering the period from 1979 to 2007, examining the relationship between per capita real life insurance premiums and per capita real GDP. Their findings indicated that developments in the life insurance market and economic growth exhibit bidirectional causality in both the long and short term. Ref. [26] utilized data from 10 OECD countries for the period 1979–2006 to investigate whether insurance sector activities support economic growth. The results showed the presence of both unidirectional and bidirectional causality relationships between economic growth and the insurance market across different OECD countries. In another study conducted by [27] the causality relationships among insurance sector penetration, stock market value, and economic growth were examined in ASEAN countries. The findings indicated the existence of bidirectional causality between insurance penetration and economic growth.

3. Materials and Methods

In this study, annual data for the G-20 countries (Australia, France, Germany, Italy, Japan, Korea, Mexico, Turkey, the United Kingdom, and the United States) covering the period from 1992 to 2022 were utilized. These ten countries have been identified by the World Bank as being among the largest 20 economies globally, underscoring their substantial impact on global economic dynamics [28]. The data of GDP, FDI, INF, GFCF variables in the study were obtained from the World Bank database, and the data of insurance variables were obtained from OECD database. All variables have been transformed into logarithmic form, except for the FDI and INF variables. Because the FDI and INF variables have been used as percentage rates. The definitions of the variables are shown in

Table 1. Variables definitions.

Variables	Definition	Source
Lngdp	GDP per capita constant 2015 (in logarithm form)	World Bank
Fdi	Foreign direct investment, net inflows (% of GDP)	World Bank
Inf	Inflation Rate	World Bank
Lngfcf	Gross fixed capital formation (constant 2015 US$)	World Bank
Lndnlife	Life Insurance density (in logarithm form)	OECD
Lndnnolife	Nonlife Insurance Density (in logarithm form)	OECD
Lndnt	Total Insurance Density (in logarithm form)	OECD
Lnprmlife	Life Insurance premium (in logarithm form)	OECD
Lnprmnolife	Nonlife Insurance premium (in logarithm form)	OECD
Lnprmt	Total Insurance premium (in logarithm form)	OECD

Note 1: All monetary measures are in real US dollars. Note 2: Insurance density, defined as the premium per capita, takes population into consideration, but neglects economic development.

.

At this stage of the study, econometric explations of these methods are presented. This study examines the long-term cointegration and causality relationship between economic growth and insurance activities. In this study is developed three distinct analytical models. The first model provides a comprehensive assessment of the insurance sector, examining the relationship between the insurance industry and economic growth through the variables Lndnt and Lnprmt. Additionally, in accordance with the frameworks established by Arena [12,29] the investigation extends to the relationship between the insurance sector and economic growth, differentiated by life and non-life insurance categories. This approach leads to the formulation of the second and third models. The second model specifically analyzes the impact of life insurance activities on economic growth, while the third model evaluates the effect of non-life insurance activities on economic growth. In this context, the model presented below has been tested to assess this relationship using panel data.

Model 1;

$$\mathit{ln}(gd{p}_t)={\alpha }_{1i}+{\beta }_{1i}\mathit{ln}d n{t}_{it}+{\beta }_{2i}\mathit{ln}p rm{t}_{it}+{\beta }_{3i} fd{i}_{it}+{\beta }_{3i} {\mathit{inf}}_{\mathit{it}}+{\beta }_{4i}\mathit{ln}g fc{f}_{it}+{\epsilon }_{it}$$

(1)

Model 2;

$$\mathit{ln}(gd{p}_t)={\alpha }_{1i}+{\beta }_{1i}\mathit{ln}d nlif{e}_{it}+{\beta }_{2i}\mathit{ln}p rmlif{e}_{it}+{\beta }_{3i} fd{i}_{it}+{\beta }_{4i} {\mathit{inf}}_{\mathit{it}}+{\beta }_{5i} \mathit{ln}g fc{f}_{it}+{\epsilon }_{it}$$

(2)

Model 3:

$$\mathit{ln}(gd{p}_t)={\alpha }_{1i}+{\beta }_{1i}\mathit{ln}dn nolif{e}_{it}+{\beta }_{2i}\mathit{ln}p rmnolif{e}_{it}+{\beta }_{3i} fd{i}_{it}+{\beta }_{4i} {\mathit{inf}}_{\mathit{it}}+{\beta }_{5i} \mathit{ln}g fc{f}_{it}+{\epsilon }_{it}$$

(3)

Before using the panel cointegration and panel causality tests, unit root test were carried out in the study. The fundamental assumption of cointegration analysis is that the first differences of the panel series are stationary. Therefore, prior to conducting the cointegration analysis, it is essential to ascertain the stationarity level of the series. In this context, the panel unit root tests employed are presented in detail below.

Levin, Lin and Chu (L-L-C, 2002) Panel Unit Root Test;

The conventional ADF test for single-equation is based on the following regression equation:

$$\Delta X_{it}={\alpha }_i+{\beta }_i{X}_{i,t-1}+{\gamma }_{i}t+{\displaystyle \sum _{j=1}^k{\theta }_{ij}\Delta X_{i,t-j}+{\epsilon }_{it}}$$

(4)

where $\Delta$ is the first difference operator, $X_{it}$ is the stock prices and dividends, ${\epsilon }_{it}$ is a white-noise disturbance with a variance of ${\sigma }^2$, and t = 1, 2, …, T indexes time. The unit root null hypothesis of ${\beta }_i=0$ is tested against the one-side alternative hypothesis of ${\beta }_i<0$, which corresponds to $X_{it}$ being stationary. The test is based on the test statistic $t_{{\beta }_i}={\widehat{\beta }}_i/se({\widehat{\beta }}_i)$ (where ${\widehat{\beta }}_i$ is the OLS estimate of ${\beta }_i$ in Equation (4) and $se({\widehat{\beta }}_i)$ is its standard error) since the single-equation ADF test may have low power when the data are generated by a near-unit-root but stationary process. Ref. [30] found that the panel approach substantially increases power in finite samples when compared with the single-equation ADF test, proposed a panel-based version of Equation (4) that restricts ${\widehat{\beta }}_i$ by keeping it identical across cross-industries as follows:

$$\Delta X_{it}={\alpha }_i+\beta X_{i,t-1}+{\gamma }_{i}t+{\displaystyle \sum _{j=1}^k{\theta }_{ij}\Delta X_{i,t-j}+{\epsilon }_{it}}$$

(5)

where i = 1, 2, …N indexes across cross-industries. Levin-Lin-Chu tested the null hypothesis of ${\beta }_1={\beta }_2=\dots .=\beta =0$ against the alternative of ${\beta }_1={\beta }_2=\dots .=\beta <0$, with the test based on the test statistic $t_{\beta }=\widehat{\beta }/se(\widehat{\beta })$ (where $\widehat{\beta }$ is the OLS estimate of $\beta$ in Equation (5), and $se(\widehat{\beta })$ is its standard error).

Im, Pesaran and Shin (IPS, 2003) Panel Unit Root Test;

Ref. [31] relaxed the assumption of the identical first-order autoregressive coefficients of the Levin-Lin-Chu test and developed a panel-based unit root test that allows $\beta$ to vary across regions under the alternative hypothesis. In addition, Im-Pesaran-Shin tested the null hypothesis of ${\beta }_1={\beta }_2=\dots .=0$ against the alternative of ${\beta }_i<0,$ for some $i$.

The Im-Pesaran-Shin test is based on the mean group approach. They use the average of the $t_{{\beta }_i}$ statistics from Equation (5) to perform the following t-bar statistic:

$$\overline{Z}=\sqrt{N}[\overline{t}-E(\overline{t})]/\sqrt{Var(\overline{t})}$$

(6)

where $\overline{t}=(1/N){\displaystyle \sum _{i=1}^N{t}_{{\beta }_i}}$, $E(\overline{t})$ and $Var(\overline{t})$ are respectively the mean and variance of each $t_{{\beta }_i}$ statistic, and they are generated by simulations [31]. This $\overline{Z}$ converges a standard normal distribution. Based on Monte Carlo experiment results, ref. [31] demonstrated their test is even more powerful than the Levin-Lin-Chu panel test in finite samples. Subsequent to the implementation of the panel unit root tests, the specifications of the panel cointegration tests utilized in this study are outlined below.

We shall apply Pedroni’s cointegration test methodology (1995, 1997 and 1999) and Panel Kao cointegration test methodology (1999) to analyse the relationship between economic growth and insurance. Ref. [32] studied the properties of spurious regressions and tests for cointegration in heterogeneous panels and derived appropriate distributions for these cases. These methods allow us to test the existence of long-term equilibrium in multivariate panels while also permitting the individual members of the dynamic and even long-term cointegration vectors to be heterogeneous.

Like the IPS panel unit root test, the panel cointegration tests proposed by Pedroni also take heterogeneity into account, utilizing specific parameters that allow for variability among individual members. Refs. [33,34] derived the asymptotic distributions of seven different statistics and examined the small sample performance of panel data cointegration tests. Four of these seven statistics are based on the principle of cointegration along the ‘Panel’ or internal dimension, while the last three are defined based on the ‘Group’ or external dimension. These different statistics are based on a model that assumes cointegration relationships among individual members are heterogeneous and are defined as;

For the Within statistics

$$Z_{\rho }^w={({\displaystyle \sum _{i=1}^N} {\displaystyle \sum _{t=1}^T} L_{11i}^{-2}{{\displaystyle \widehat{e}}}_{it-1}^2)}^{-1}{\displaystyle \sum _{i=1}^N} {\displaystyle \sum _{t=1}^T} L_{11i}^{-2}({{\displaystyle \widehat{e}}}_{it-1}\Delta {{\displaystyle \widehat{e}}}_{it}-{{\displaystyle \widehat{\lambda }}}_i):Panel Rho\_stat$$

$$Z_t^w={({{\displaystyle \tilde{s}}}_{NT}^{\ast 2}{\displaystyle \sum _{i=1}^N} {\displaystyle \sum _{t=1}^T} L_{11i}^{-2}{{\displaystyle \widehat{e}}}_{it-1}^{\ast 2})}^{-1/2}{\displaystyle \sum _{i=1}^N} {\displaystyle \sum _{t=1}^T} L_{11i}^{-2}({{\displaystyle \widehat{e}}}_{it-1}^{\ast }\Delta {{\displaystyle \widehat{e}}}_{it}^{\ast }):Panel Adf\_stat$$

$$Z_{pp}^w={({{\displaystyle \tilde{\sigma }}}^2{\displaystyle \sum _{i=1}^N} {\displaystyle \sum _{t=1}^T} L_{11i}^{-2}{{\displaystyle \widehat{e}}}_{it-1}^2)}^{-1/2}{\displaystyle \sum _{i=1}^N} {\displaystyle \sum _{t=1}^T} L_{11i}^{-2}({{\displaystyle \widehat{e}}}_{it-1}\Delta {{\displaystyle \widehat{e}}}_{it}-{{\displaystyle \widehat{\lambda }}}_i):Panel PP\_stat$$

$$Z_v^w={({\displaystyle \sum _{i=1}^N} {\displaystyle \sum _{t=1}^T} L_{11i}^{-2}{{\displaystyle \widehat{e}}}_{it-1}^2)}^{-1}:Panel V\_stat$$

For the Between statistics

$$Z_{\rho }^B={\displaystyle \sum _{i=1}^N} {({\displaystyle \sum _{t=1}^T} {{\displaystyle \widehat{e}}}_{i,t-1}^2)}^{-1}{\displaystyle \sum _{t=1}^T} ({{\displaystyle \widehat{e}}}_{it-1}\Delta {{\displaystyle \widehat{e}}}_{it}-{{\displaystyle \widehat{\lambda }}}_i):Group Rho\_stat$$

$$Z_t^B={\displaystyle \sum _{i=1}^N} {({{\displaystyle \widehat{\sigma }}}_i^2{\displaystyle \sum _{t=1}^T} {{\displaystyle \widehat{e}}}_{i,t-1}^2)}^{-1}{\displaystyle \sum _{t=1}^T} (({{\displaystyle \widehat{e}}}_{it-1}\Delta {{\displaystyle \widehat{e}}}_{it}-{{\displaystyle \widehat{\lambda }}}_i):Group Adf\_stat$$

$$Z_{pp}^B={\displaystyle \sum _{i=1}^N} {{\displaystyle \left({\displaystyle \sum _{t=1}^T} {{\displaystyle \widehat{s}}}^{\ast 2}{{\displaystyle \widehat{e}}}_{it-1}^{\ast 2}\right)}}^{-1}{\displaystyle \sum _{t=1}^T} ({{\displaystyle \widehat{e}}}_{it-1}^{\ast }\Delta {{\displaystyle \widehat{e}}}_{it}^{\ast }):Group PP\_stat$$

with,

$$\widehat{\lambda }=\frac{1}{T}{\displaystyle \sum _{s=1}^{k_i}} (1-\frac{s}{k_i+1}){\displaystyle \sum _{t=s+1}^t} {{\displaystyle \widehat{\mu }}}_{it}{{\displaystyle \widehat{\mu }}}_{it-s}$$

$${{\displaystyle \widehat{s}}}_i^2=\frac{1}{T}{\displaystyle \sum _{t=s+1}^t} {{\displaystyle \widehat{\mu }}}_{it}^2{{\displaystyle \widehat{\sigma }}}^2=s_i^2+2{{\displaystyle \widehat{\lambda }}}_i$$

$${{\displaystyle \widehat{\sigma }}}^2=s_i^2+2{{\displaystyle \widehat{\lambda }}}_i$$

$${{\displaystyle \tilde{\sigma }}}_{NT}^2\frac{1}{T}{\displaystyle \sum _{i=1}^N} {{\displaystyle \widehat{L}}}_{11i}^{-2}{{\displaystyle \widehat{\sigma }}}_i^2$$

$${{\displaystyle \widehat{s}}}_i^{\ast 2}=\frac{1}{T}{\displaystyle \sum _{t=s+1}^t} {{\displaystyle \widehat{\mu }}}_{it}^{\ast 2}{{\displaystyle \tilde{s}}}_{NT}^{\ast 2}=\frac{1}{T}{\displaystyle \sum _{t=s+1}^t} {{\displaystyle \widehat{s}}}_{it}^{\ast 2},{{\displaystyle \widehat{L}}}_{11i}^2{\displaystyle \sum _{t=1}^T} {{\displaystyle \widehat{\eta }}}_{it}^2+\frac{2}{T}{\displaystyle \sum _{s=1}^{k_i}} (1-\frac{s}{k_i+1}){\displaystyle \sum _{i=1}^T} {{\displaystyle \widehat{\eta }}}_{it}{{\displaystyle \widehat{\eta }}}_{it-s}$$

and where the residuals are extracted from the above regressions:

$${{\displaystyle \widehat{e}}}_{it}=\widehat{\rho }{{\displaystyle \widehat{e}}}_{it-1}+{{\displaystyle \widehat{u}}}_{it}$$

$${{\displaystyle \widehat{e}}}_{it}=\widehat{\rho }{{\displaystyle \widehat{e}}}_{it-1}+{\displaystyle \sum _{k=1}^{K_i}} {{\displaystyle \widehat{\gamma }}}_{ik}\Delta {{\displaystyle \widehat{e}}}_{it-k}+{{\displaystyle \widehat{u}}}_{it}$$

$$\Delta y_{it}={\displaystyle \sum _{m=1}^M} {{\displaystyle \widehat{b}}}_{mi}\Delta X_{mit}+{{\displaystyle \widehat{\eta }}}_{it}$$

Note that in the above writings $L_i$ represents the 1th component of the Cholesky decomposition of the residual Variance-Covariance matrix, $\widehat{\lambda }$ and ${{\displaystyle \tilde{\sigma }}}_{NT}^2$ are two parameters used to adjust the autocorrelation in the model, ${\sigma }_i$ and s_i² are the contemporaneous and long-run individual variances.

$${\displaystyle \frac{{\chi }_{NT}-\mu \sqrt{N}}{\sqrt{v}}}\to N(\mathrm{0,1})$$

where ${\chi }_{NT}$ is the statistic under consideration among the seven proposed, N and T are the sample parameter values and $\mu$ and $\nu$ are parameters tabulated in [34].

Ref. [33] demonstrated that for values of T greater than 100, all seven proposed statistics perform well and are quite stable. However, for smaller samples (when T is less than 20), the Group ADF-Statistic (non-parametric) is the most powerful, followed by the Panel v-Statistic and the Panel rho-Statistic. Therefore, in our study on panel cointegration testing, only the group ADF-statistic will be considered. The finite sample distributions for the seven statistics were tabulated by [33] through Monte Carlo simulations. The calculated test statistics must be greater (in absolute value) than the tabulated critical value to reject the null hypothesis of no cointegration.

Ref. [35] proposed two sets of specifications for the DF test statistics. The first set depends on consistent estimation of the long-run parameters, while the second one does not. Under the null hypothesis of no cointegration, the residual series $e_{it}$ should be non-stationary. The model has varying intercepts across the cross-sections (the fixed effects specification) and common slopes across i. The DF test can be calculated from the estimated residuals as:

$${\widehat{e}}_{it}=\rho {\widehat{e}}_{it-1}+v_{it}$$

The null hypothesis of non-stationarity can be written as $H_0: \rho =1$. Kao constructed new statistics whose limiting distributions, $N\left(0,1\right)$, are not dependent on the nuisance parameters, that are called $D{F}_{\rho }^{*}$ and $D{F}_t^{*}$ (where it is assumed that both regressors and errors are endogenous). Alternatively, he defines a bias-corrected serial correlation coefficient estimates and, consequently, the bias-corrected test statistics and calls them $D{F}_{\rho }$ and $D{F}_t$. In this case, the assumption is the strong exogeneity regressors and the errors. Finally, ref. [35] also proposed an ADF type of regression and an associated ADF statistic.

Finally, the study aims to explore the short-run dynamic bivariate panel causality among the variables using the model. Ref. [36] suggested a simple approach for testing the null hypothesis of homogeneous non-causality against the alternative hypothesis of heterogeneous non-causality. This test has to be applied to a stationary data series using the fixed coefficients in a vector autoregressive (VAR) framework. The significance of this test is that it allows for having a different lag structure and also heterogeneous unrestricted coefficients across the cross-sections under both the hypotheses. Under the null hypothesis, no causality in any cross-section is tested against the alternative hypothesis of causality at least for a few cross-sections.

4. Empirical Findings

This section presents empirical findings that reveal the relationship between the insurance sector and economic growth. The study aims to examine the relationship between growth in the insurance sector and economic development. In this context, a group of countries with significant economic importance, the G-20 economies, has been selected, and the countries with acces-sible data, such as Australia, France, Germany, Italy, South Korea, Japan, Mexico, Turkey, the United Kingdom, and the United States, have been included in the study. The annual GDP and some insurance data of these 10 countries for the period between 1992 and 2022 have been used in the study. The data used have been obtained from the World Bank database and the OECD database. The details of the variables considered in the study are provided in

Table 2. Descriptive Statistics.

Variables	Lngdp	Lndn Life	Lndn Nolife	Lndnt	Lnprm Life	Lnprm Nolife	Lnprmt	Inf	Fdi	Lngfcf
Mean	28.36	6.53	6.38	7.24	10.9	10.85	11.66	6.3	1.88	3.13
Median	28.28	7.24	6.69	7.8	11.42	10.93	12	2.03	1.53	3.1
Maximum	30.69	9.04	8.66	9.21	14.18	14.54	15.07	143.64	12.73	3.62
Minimum	26.45	0.78	2.76	2.89	4.87	6.86	6.99	−1.88	−3.61	2.75
Std. Dev.	0.92	1.82	1.27	1.47	1.95	1.58	1.68	16.37	1.85	0.18
Skewness	0.66	−1.44	−1.01	−1.31	−0.99	0.02	−0.52	5.03	2.32	0.34
Kurtosis	3.43	3.93	3.27	3.55	3.56	3.14	3.08	31.58	12.2	2.59

.

In the study, prior to testing for cointegration and causality relationships among the variables, several panel unit root tests were conducted to determine the stationarity levels of the variables. In this context, commonly used unit root tests in panel data models, such as [30,31] were employed. The statistics obtained from the application of these unit root tests are presented in

Table 3. Level (I0) Panel Unit Root Test Results.

Level: I(O)	LLC		IPS
Variable	Constant	Cons + Trend	Constant	Cons + Trend
	t ist./prob	t ist./prob	t ist./prob	t ist./prob
lngdp	−5.769	−2.311	−1.444	0.508
	0.000 ***	0.010 **	0.074 *	0.694
Inf	−1.292	1.586	57.766	−1.033
	0.098 *	0.944	0.000 ***	0.151
Fdi	−6.0 85	−0.127	−6.145	−1.805
	0.000 ***	0.450	0.000 ***	0.036 **
Lngfcf	−1.325	1.693	−1.705	0.294
	0.093 *	0.955	0.044 **	0.616
Lndnt	−3.029	0.557	−1.965	1.049
	0.001 ***	0.711	0.025 **	0.853
Lndnlife	−6.021	−0.840	−3.338	−0.046
	0.000 ***	0.200	0.000 ***	0.482
Lndnnolife	0.746	1.220	1.913	0.111
	0.772	0.889	0.972	0.544
Lnprmt	−2.660	−1.214	0.111	0.487
	0.004 ***	0.112	0.544	0.687
Lnprmlife	−3.494	−1.277	−0.842	0.377
	0.000 ***	0.101	0.200	0.647
Lnprmnolife	−0.874	−0.403	1.345	−0.729
	0.191	0.344	0.911	0.233

Note: *** indicates significance at the 1% level, ** at the 5% level, and * at the 10% level.

and

Table 4. First Difference (I1) Panel Unit Root Test Results.

Level: I(1)	LLC		IPS
Variable	Constant	Cons + Trend	Constant	Cons + Trend
	t ist./prob	t ist./prob	t ist./prob	t ist./prob
lngdp	−12.512	−9.343	−11.143	−10.426
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Inf	−17.729	−16.605	−16.541	−16.333
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Fdi	−9.999	−3.442	−13.492	−9.660
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Lngfcf	−10.021	−9.322	−8.648	−8.663
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Lndnt	−9.112	−7.460	−8.424	−8.084
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Lndnlife	−7.848	−7.923	−7.462	−8.493
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Lndenolife	−10.924	−8.366	−10.057	−8.257
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Lnprmt	−9.706	−7.073	−9.356	−8.387
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Lnprmlife	−6.310	−6.705	−6.973	−8.198
	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Lnprmnolife	−9.609	−5.834	−9.879	−7.378
	0.000 ***	0.000 ***	0.000 ***	0.000 ***

Note: *** indicates significance at the 1% level.

.

The results of the conducted unit root tests revealed that while some variables exhibited stationarity at level, others did not. Consequently, it was decided to examine the stationarity of the data for the variables in first differences (I(1)). The findings indicate that all variables representing both economic development and the insurance sector, as well as other control variables, demonstrate stationarity in first differences.

Following the determination that non-stationary variables are stationary in first differences, the existence of a long-term cointegrated relationship between economic development and other variables was examined using the Pedroni and Kao panel cointegration tests. The Pedroni cointegration test allows for both dynamic and fixed effects to differ across the panel’s cross-sections, as well as permitting the cointegrating vector to vary among these cross-sections under the alternative hypothesis. Some favorable characteristics of Pedroni’s approach emerge in the context of the null hypotheses of no trend and no cointegration derived from the approaches of [37]. Notably, the Pedroni tests allow for multiple explanatory variables, the diversification of the cointegration vector across different parts of the panel, and also accommodate heterogeneity of errors across cross-sectional units, as highlighted by [38]. To capture within and between effects in the panel, seven different cointegration tests are presented. Another cointegration test to be utilized in the study is the one introduced by [35] which employs DF and ADF tests for panel data analysis.

The findings obtained from the conducted cointegration tests indicate the existence of a long-term cointegrated relationship between economic growth and the insurance sector. Accordingly, although the Panel rho-Statistic values for the first model are not statistically significant, the Panel v, Panel PP, and Panel ADF values are statistically significant, leading to the rejection of the null hypothesis (there is no cointegration relationship among the series). Additionally, it has been determined that the Group PP and Group ADF values are also statistically significant, further confirming the cointegration relationship. A similar situation has been found to hold for the second and third models constructed. These results are also supported by the findings from the Kao cointegration test. The statistical values and significance levels obtained from the cointegration tests are detailed in

Table 5. Pedroni Panel Cointegration Results.

Within Dimension Based Test
Variables		1		2		3
		Statistic	Prob.	Statistic	Prob.	Statistic	Prob.
Panel v-Statistic		4.331939	0.0000 ***	−0.87675	0.8097	−0.243394	0.5961
Panel rho-Statistic		1.415767	0.9216	0.612724	0.7300	1.249428	0.8942
Panel PP-Statistic		1.746264	0.0404 **	−2.664018	0.0039 **	−1.698207	0.0447 **
Panel ADF-Statistic		−1.434402	0.0757 *	−2.923054	0.0017 **	−2.125353	0.0168 **
Between Dimension Based Test
Group rho-Statistic		2.571126	0.9949	2.565674	0.9949	2.446048	0.9928
Group PP-Statistic		−1.299222	0.0969*	−1.983184	0.0237 **	−2.014765	0.0220 **
Group ADF-Statistic	−2.58082	0.0049 **	−1.996417	0.0229 **	−3.687855		0.0001 ***
Kao Residual Cointegration Test
ADF Statistic	3.361349	0.0004 ***	1.316778	0.0940 *	1.894158		0.0291 **

Note: *** indicates significance at the 1% level, ** at the 5% level, and * at the 10% level.

.

The cointegration tests have identified the existence of a long-term relationship between economic growth and the insurance sector. Following the determination of this long-term relationship, the magnitude of the coefficients related to this relationship was examined. Specifically, the effects of the insurance variables included in the study, representing the insurance sector, along with other control variables on economic growth were analyzed using the DOLS (Dynamic Ordinary Least Squares) and FMOLS (Fully Modified Ordinary Least Squares) methods developed by [39,40] While findings from both methods are reported in the study, it is noted that the FMOLS approach demonstrates greater consistency in long-term parameter estimation compared to DOLS.

Below examining the results of the Panel FMOLS test presented in

Table 6. Panel FMOLS Results.

Variable	1		2		3
	Coefficient	Prob.	Coefficient	Prob.	Coefficient	Prob.
Lndnt	0.0606	0.0006 ***	-	-	-	-
Lnprmt	0.0213	0.3024	-	-	-	-
Lndnlife	-	-	0.0233	0.7412	-	-
Lnprmlife	-	-	0.0471	0.5203	-	-
Lndnnlife	-	-	-	-	−0.0047	0.8484
Lnprmnlife	-	-	-	-	0.0617	0.0111 **
Inf	0.0008	0.0000 ***	0.0007	0.0209 **	0.0065	0.0000 ***
Fdi	0.0044	0.0000 ***	0.0047	0.0229 **	0.0006	0.0000 ***
Lngfcf	0.1387	0.0000 ***	0.1321	0.0040 ***	0.1575	0.0000 ***
R-squared	0.9988		0.9988		0.9986

Note: *** indicates significance at the 1% level, and ** at the 5% level.

, it has been determined that the total insurance density (lndnt) and total gross premium (lnprmt) variables, representing the insurance sector in the first model, have a positive effect on economic growth, with the coefficient for lndnt being 0.0606, which is statistically significant at the 1% level. The study aimed to analyze the impact of the insurance sector separately in terms of life and non-life insurance. Consequently, FMOLS results for Model 2, which examines the effects of life insurance, and Model 3, which focuses on non-life insurance, are included. In Model 2, the test statistics calculated for the variables representing life insurance, Lndlife and Lnprmlife, were found to be statistically insignificant.

Therefore, no conclusions could be drawn regarding the impact of life insurance on economic growth. On the other hand, it was determined that the non-life insurance type examined in Model 3 positively affects economic growth. Specifically, the coefficient for the Lnprmnlife variable representing non-life insurance was calculated to be 1.4820, which is statistically significant at the 1% level. However, the coefficient for another variable representing non-life insurance, Lndnnlife, was found to be statistically insignificant. Moreover, although it was concluded that the Inf and Fdi variables also influence economic growth, the calculated coefficient values for these variables were observed to be quite low. The Lngcf variable positively affected economic growth with coefficient values of 0.1387 for Model 1, 0.1321 for Model 2, and 0.1575 for Model 3.

Upon examining the Panel DOLS results presented in

Table 7. Panel DOLS Results.

Variable	1		2		3
	Coefficient	Prob.	Coefficient	Prob.	Coefficient	Prob.
Lndnt	0.1162	0.0626 *	-	-	-	-
Lnprmt	−0.0354	0.6181	-	-	-	-
Lndnlife	-	-	0.0151	0.8473	-	-
Lnprmlife	-	-	0.0470	0.5667	-	-
Lndnnlife	-	-	-	-	−0.6265	0.1485
Lnprmnlife	-	-	-	-	0.1482	0.0000 ***
Inf	0.0015	0.0885 *	0.0007	0.0410 **	0.1072	0.2346
Fdi	0.0065	0.1367	0.0030	0.1871	0.0272	0.0398 **
Lngfcf	0.1726	0.0159 **	0.1783	0.0004 ***	0.1524	0.0000 ***
R-squared	0.9998		0.9987		0.6745

Note: *** indicates significance at the 1% level, ** at the 5% level, and * at the 10% level.

, it is observed that the insurance sector has a positive effect on economic growth, similar to the findings from the FMOLS results. Specifically, for Model 1, the coefficient for the Lndnt variable indicates a value of 0.1162, suggesting that the insurance sector as a whole positively influences economic growth. However, when the effects of the insurance sector are analyzed in two different models concerning life and non-life insurance types, it is determined that only the non-life insurance type in Model 3 has a positive effect on economic growth. In Model 3, the coefficient for the lnprmnlife variable is calculated to be 0.1482, which is statistically significant at the 1% level. Additionally, the Lngfcf variable positively affects economic growth in all three models, with coefficient values of 0.1726, 0.1871, and 0.1524, respectively.

In line with the study’s objectives, the final stage examines the existence of a causal relationship among the series. In this context, the causality test developed by [36] is utilized. This method accounts for cross-sectional dependence and heterogeneity among countries in the panel, addresses situations where the time dimension exceeds the cross-sectional dimension (N), and provides reliable results even in unstable panel data sets [36]. The findings from the conducted causality test are presented in

Table 8. Dimitrescu Hurlin Causality Test Results.

Causality to LNGDP			Causality from LNGDP
Null Hypothesis:	Stat.	Prob.	Null Hypothesis:	Stat.	Prob.
Lndnt ⟹ Lngdp	2.8079	0.0949 *	Lngdp ⟹ Lndnt	0.5177	0.4724
Lnprmt ⟹ Lngdp	1.6323	0.2024	Lngdp ⟹ Lnprmt	4.7139	0.0307 **
Lndnlife ⟹ Lngdp	0.5061	0.4774	Lngdp ⟹ Lndnlife	0.0348	0.8521
Lnprmlife ⟹ Lngdp	0.0457	0.8307	Lngdp ⟹ Lnprmlife	1.4913	0.223
Lndnnlife ⟹Lngdp	5.2625	0.0225 **	Lngdp ⟹ Lndnnlife	0.6461	0.4222
Lnprmnlife ⟹Lngdp	3.9631	0.0474 **	Lngdp ⟹ Lnprmnlife	2.9355	0.0877 *
Fdi ⟹ Lngdp	0.0050	0.9434	Lngdp ⟹ Fdi	0.0515	0.8205
Inf ⟹ Lngdp	0.0268	0.8700	Lngdp ⟹ Inf	2.2842	0.1318
Lngcf ⟹ Lngdp	4.4824	0.0000 ***	Lngdp ⟹ Lngcf	1.9437	0.1643

Note: *** indicates significance at the 1% level, ** at the 5% level, and * at the 10% level.

.

Table 8 presents the results of the causality test conducted in two directions: from the independent variables to the economic growth variable, represented by Lngdp, which is considered the dependent variable in this study, and vice versa. According to the results obtained, there is a causal relationship from the Lndnt variable, which represents the insurance sector as a whole, to Lngdp, and this relationship is statistically significant at the 10% level. A similar situation has been found to hold for the non-life insurance type. Specifically, a causality exists from the variables representing non-life insurance, Lndnnlife and Lnprmnlife, to the economic growth variable, Lngdp, and this relationship is statistically significant. However, no causality could be identified from the variables representing life insurance to Lngdp. Additionally, it is noteworthy that a causal relationship exists from the Lngcf variable to Lngdp, apart from the insurance sector variables. When examining the causality from the dependent variable Lngdp to the other independent variables, it is determined that there is a causality from Lngdp to Lnprmt and Lnprmnlife. These results indicate a bidirectional causality relationship between insurance premiums and GDP.

5. Conclusions and Implications

The insurance sector plays a critical role in economic growth by ensuring financial stability, promoting investments, and supporting sustainable development initiatives through risk management. This study aims to reveal the connection between insurance sector activities and economic growth in G-20 countries. Previous research has indicated a complex relationship between the development of the insurance sector and economic growth, yielding varied results.

In this context, several G-20 countries, such as Australia, France, Germany, Italy, South Korea, Mexico, Japan, Turkey, the United Kingdom, and the United States where data can be reliably accessed—have been included in the study. The relationship between economic growth and the insurance sector is examined using various data related to the insurance sector alongside GDP data from these economically significant countries. Three different models have been constructed in this study. In the first model, the insurance sector is analyzed holistically, investigating the relationship between the insurance sector and economic growth through the Lndnt and Lnprmt variables. Additionally, the relationship between the insurance sector and economic growth is examined based on life and non-life insurance types, leading to the creation of the second and third models. Accordingly, the second model tests the relationship between life insurance activities and economic growth, while the third model explores the relationship between non-life insurance activities and economic growth. The study employs the Pedroni and Kao cointegration tests, FMOLS and DOLS estimators, and finally, the causality test developed by [36].

The findings obtained from the Pedroni and Kao cointegration tests indicate the existence of a long-term cointegrated relationship between economic growth and the insurance sector. Specifically, for the first model, it is observed that the Panel v, Panel PP, Panel ADF, Group PP, and Group ADF values are statistically significant, leading to the rejection of the null hypothesis (no cointegration relationship among the series). A similar situation is found to be valid for the second and third models as well, and these results are further supported by the Kao cointegration test. Following the determination of a long-term relationship between economic growth and the insurance sector through the cointegration tests, the magnitude of the coefficients related to this relationship was examined using FMOLS and DOLS estimators. Initially, when reviewing the panel FMOLS test results, it was determined that in the first model, which considers the insurance sector as a whole, the Lndnt variable has a coefficient value of 0.0606, indicating a positive effect on economic growth, and this result is statistically significant at the 1% level. Additionally, the study aimed to examine the effect of the insurance sector in two separate dimensions, focusing on life and non-life insurance types. The FMOLS results for Model 2, which addresses the impact of life insurance, and Model 3, which examines the impact of non-life insurance, are also presented. In Model 2, concerning life insurance, it was found that the coefficient values calculated for both the Lndlife and Lnprmlife variables are not statistically significant. Therefore, no conclusions could be drawn regarding any impact of life insurance on economic growth. Conversely, in Model 3, it was observed that non-life insurance positively affects economic growth. Furthermore, the control variable Lngcf was found to positively influence economic growth with coefficient values of 0.1387 for Model 1, 0.1321 for Model 2, and 0.1575 for Model 3.

Upon examining the Panel DOLS results, findings similar to those of the FMOLS were obtained, indicating that the insurance sector has a positive effect on economic growth. Specifically, for Model 1, the Lndnt variable has a coefficient value of 0.1162, demonstrating that the insurance sector as a whole positively influences economic growth. However, when the effects of the insurance sector are analyzed within two different models focusing on life and non-life insurance types, it is found that only in Model 3, which addresses non-life insurance, does the insurance sector have a positive impact on economic growth. The Lngcf variable was identified to positively affect economic growth with coefficient values of 0.1726, 0.1871, and 0.1524 across all three models. In the final phase of the study, the existence of a causal connection among the series is examined using the [36] causality test. The results indicate a causal relationship from the Lndnt variable, representing the total insurance sector, to Lngdp, which is statistically significant at the 10% level. A similar situation is found to be valid for non-life insurance types as well. Specifically, a causality is established from the Lndnnlife and Lnprmnlife variables, representing non-life insurance, to Lngdp, and this relationship is statistically significant. Conversely, no causality could be determined from the variables representing life insurance to Lngdp. Additionally, a noteworthy causal relationship exists from the Lngcf variable to Lngdp, apart from the insurance sector variables. When examining the causality from the dependent variable Lngdp to other independent variables, it was found that there is a causality from Lngdpi to Lnprmt and Lnprmnlife. These results particularly indicate a bidirectional causality between insurance premiums and economic growth. The findings of this study support both supply-leading and demand-following theories, aligning with the works of [23,24,25,26,27].

The outputs obtained from this study provide evidence that the development of the insurance sector plays a significant role in the economic advancement of countries with strong economies, such as those in the G-20. A well-developed insurance industry enables countries to diversify their investment resources, facilitate capital accumulation, and enhance economic resilience against natural disasters. In this regard, whether in G-20 economies or other developed or developing economies, there is a recognized benefit in increasing the steps taken to promote the development of the insurance sector to achieve sustainable economic growth. It is believed that policymakers should focus on the relationship between insurance activities and economic growth while considering the unique dynamics and characteristics of each country when formulating policies.