“Net investment position and the stock market: The case of traditional and ESG indices”

This paper explores the influence of traditional and ESG stock market indices on a country’s net international investment position. To do this, different methods, including ANOVA analysis, multiply regression analysis, correlation analysis, VAR-analysis and R/S-analysis, as well as the Granger causality test, are applied to quarterly data on the net international investment position, traditional and ESG indices from Finland, Sweden, France, Spain and Ukraine over the period 2005–2021. The results of descriptive statistics show that ESG indices are more volatile than traditional, but these differ- ences are statistically insignificant according to ANOVA analysis. Correlation analysis provides direct evidence that ESG indices are highly correlated with their traditional analogues (correlation level varies from 0.88 to 0.96). Regression analysis results show that traditional and ESG stock market indices have no significant impact on the net international investment position. ESG stock market indices and net international investment position data are persistent, and autoregressive models can be applied to these data sets. On average, Hurst exponent is above 0.75 for the case of ESG indices and above 0.85 for the net investment position. This paper provides recommendations to improve the responsible investment framework.


INTRODUCTION
The COVID-19 pandemic has reduced total global investment flows by $0.6 trillion annually and foreign direct investment by 40% (UNCTAD, 2020). Countries with an unstable investment position and unfavorable investment climate suffer significant losses. For example, in Ukraine, capital investment declined by 40% during the pandemic. Overall, since 2007 in Ukraine the net growth of foreign direct investment has been less than the normative value (7% of GDP).
In these conditions, developed countries are trying to compensate the deterioration of their investment position by shifting from traditional to socially responsible investment (ESG).
In particular, in the EU in the next decade as a part of EU Green Deal, it is planned to accumulate around €1 trillion from the EU budget and related mechanisms for the circular economy support, infrastructure upgrading, biodiversity, small and medium-sized enterprises, agriculture and innovation (EESC, 2019).
This investment support is primarily aimed at investing in "green" post-pandemic recovery and has the responsible investment (RI) or-igin. It aims to improve the net international investment position of EU countries through multiplier effects and intensify the transition to sustainability and achievement of SDGs.
A possible way to solve the current problems in countries like Ukraine is the use of the relevant experience from developed European countries for creating a basis for RI attraction. It can help to improve the net international investment position and accelerate sustainable development by intensifying RI and benchmarks in the stock market.
The ESG investment dynamics in a country can be illustrated by key indices of sustainable development (ESG indices), which has been accelerated recently with the intensification of responsible investing. However, responsible investment is not widely used by countries. As a result, it is hard to estimate the influence of ESG indices on investment processes.
Stock markets are major influencers on the investment climate and the country's position in capital markets (Baumohl, 2012;Thalassinos et al., 2015). Still, the relationship between the net international investment position and stock market indices is not explored yet in the academic literature.

LITERATURE REVIEW
International investment position is a statistical report that introduces the value and structure of a country's external financial assets and liabilities at a certain period (Ukrstat, 2020) and illustrates the balance of investment flows. It covers categories such as a direct investment: equity and debt instruments; portfolio investments: equity instruments; debt securities; financial derivatives; other investments: other equity instruments, currency and deposits, loans, trade credits and advance repayments; reserve assets.
Given the number of components, the investment position is a rather complex concept. Under its complexity, based on the basis of bibliometric analysis, this study proposes to carry out a review of previous research papers in the field of the impact of stock indices on the net international position. They take into account modern algorithms for finding sources by exact parameters and keywords from the largest scientometric databases. The combination of these methods (In-built Scopus instruments by Elsevier, Inbuilt Web of Science instruments by Clarivate Analytics, Publish or Рerish software) is used to select and summarize the academic background related to a country's net investment position and the impact of traditional and ESG stock market indices: • In-built Scopus instruments by Elsevier, Inbuilt Web of Science instruments by Clarivate Analytics -for selection and initial analysis of publications from Scopus and Web of Science; • Publish or Рerish -for selection and initial analysis of publications from Google Scholar; • Each search and request within the meta-analysis instruments are formed using a logic operator as of January 25, 2022 for 2000-2021.
The research queries are as follows: • Net international investment position; • Net international investment position AND stock indices; • Net international investment position AND responsible investment.
The generalization of the array of scientific papers on three scientometric bases (a total of 448 papers) indicates the prevalence of scientific papers within the query Net international investment position. The query Net international investment position AND stock indices, as well as the query Net international investment position AND responsible investment are represented by a relatively small number of publications -10 and 9, respectively.
According to the query results on search terms for 11 years, countries' net investment positions are not considered actively enough. For the most part, scientific queries have been actively started since 2010. The largest number of significant publications on this topic is concentrated in the Google Scholar database (by the number of citations, the Hirsch index). However, even in this database, neither the impact of traditional indices nor responsible investment is considerable.
The results of cluster analysis by publication keywords from the WoS and Scopus databases ( Figure  A1) confirm the conclusion. Within the predominant topic of net investment position (the green cluster), there are no keywords that would link it to stock indices or responsible investments (ESG).
Given the novelty of the topic and the lack of longterm research, the scientific explanation of the RI impact on the net international investment position is not sufficiently available. The generalized representation of interconnections between the authors who study the countries` investment positions shows a small number of such scientists (about 30) during 2000-2021, as well as the lack of significant scientific schools in this area ( Figure A2).
Academic papers in this field have a predominantly national context and consider the countries' net international investment position and some influence of stock market indices, in particular:  The RI context is mainly not used in the works mentioned above. Only in Bruna's (2013) paper, the results confirm that with such a net investment position, the Czech Republic's economy cannot meet sustainability needs, and its deterioration negatively affects sustainable development. In addition, Lisicke and Maleček (2012) investigate factors that influenced sustainability of the Czech international investment position. But the role of traditional or ESG indices is not underlined.  As can be seen, the study of the relationship between the net international investment position and stock market indices is not widely represented in the academic literature.
This paper aims to explore the impact of traditional and ESG indices on a country's net international investment position. This is quite a pioneer topic in modern academia.

DATA AND METHODOLOGY
To model the impact of traditional and ESG indices on the net international investment position, the stock markets of Ukraine (developing country), as well as Finland, Sweden, France and Spain (developed countries) are selected as analysis objects. This choice is made due to the available data for traditional and ESG indices. Three data sets are used in this paper such as a country's net investment position, ESG index, and traditional index. The sources and data periods by country are given in Table 3.
The methodology of this paper includes the following methods: • Traditional descriptive statistics is used to determine the differences in the statistical characteristics of analyzed data sets; • Variance analysis (ANOVA-analysis) is used to identify statistically significant differences between the data sets; • Correlation analysis is performed to identify how synchronous are the variables; • Granger causality test is applied to clarify the correlation analysis results, as well as to determine which of the indicators is dependent and which is independent;  • R/S data analysis is used to identify probable differences in the data sets and determine the possible predictability of data based on their previous values. In this paper, the methodology similar to Plastun et al. (2018) is applied; • Autocorrelation function analysis is performed to determine the optimal lag of autoregressive models; • Regression analysis determines the ability to predict the countries' investment position based on three models' stock exchange market dynamics.
• Model 1. The first variable is the previous value of the investment position indicator with a lag selected from the autocorrelation function analysis. The ESG and the traditional indices are used as additional variables in the model.
• Model 2. The ESG-index is a basic variable.
• Model 3. The basic variable is the traditional index. Models 2 and 3 evaluate the possibility of using ESG indices and traditional indices as a key factor influencing the investment position.
• VAR method is applied to build vector autoregressive models that describe the impact of ESG indices and traditional indices on the investment position.

RESULTS
Descriptive statistics for the first differences (Table B1) show that mostly ESG indices are more volatile than traditional ones (standard deviation, the growth rate in traditional indexes is less than in the ESG indices). Accordingly, from the risk point of view, ESG indices do not have advantages for investors. However, the average yield on ESG indices is usually higher than traditional ones. For example, in Ukraine, traditional indices are much more volatile and profitable. This can be explained by the specifics of Ukrainian data, as the traditional stock index is formed from stock prices denominated in hryvnia, and ESG index data are quoted in euros. The volatility of the hryvnia exchange rate may be a decisive factor for the recorded differences.
ANOVA analysis (Table 4) shows that there are no statistically significant differences between the data sets behavior. Thus, previous evidence of the difference between traditional and ESG indices in the context of their volatility can be considered statistically insignificant. The general conclusion from the ANOVA analysis is that a country's investment position, the ESG index and the tradi-  Note: * p-value is given in parentheses.
tional index behave very similarly, at least from the point of dynamic changes.
The correlation analysis results ( As for other countries, the results are mixed. In France, there is a strong relationship between the dynamics of ESG indices and the investment position. The correlation coefficient is also negative for the traditional index, but the correlation coefficient is more than twice lower. The weaker relation is typical for Spain, and the traditional index is more related to the investment position than the ESG index.
In Sweden, the situation is opposite. The relationship between the investment position and the indices is more robust in the traditional index and is direct.
To confirm the correlation analysis results and determine which of the indicators is the regressor or regressant, the Granger test is conducted for both the absolute values and first differences ( Table 6).
The results show that the investment position does not affect the stock market in most cases, and the stock market does not affect the investment po- R/S data analysis is vital to provide additional evidence about potential differences in the analyzed data sets, and determining the possible data predictability based on previous statistics (Table 7). It is impossible to assess the data for Ukraine due to the small size of the data set. As for Finland, Sweden, Spain and France, the investment position dynamics are characterized by strong persistence with a much lower level in the stock market. Moreover, a higher persistence level is observed on the ESG index dynamics. That is, ESG indices are more predictable than traditional ones. An autocorrelation function analysis is used to obtain detailed results (Table C1). The results show that the optimal lag for describing the investment position dynamics is 1. Since the autocorrelation function values are pretty high and statistically significant, a model for forecasting the investment position based on previous data can be used.
To determine the real possibility of investment position forecasting based on the stock market dynamics, the next step is to conduct a regression analysis. The study considers the model of the investment position dependent on many factors.  Table 8.
As shown in Table 8, all first-order autoregressive models are adequate, as proved by determination coefficients from a minimum of 0.74 for Sweden to a maximum of 0.96 for France. In particular, for most countries (except Spain), the previous value of the investment position and the stock indices dynamics have a statistically significant influence on the dependent variable. Both traditional and ESG indices do not have a statically significant effect. This is additional evidence in favor of the fact that the investment position dynamics is not VAR analysis for time series has its peculiarities, which requires taking the following steps: • checking the time series for stationarity and solving the non-stationarity problem in case of its presence; • determining the optimal number of lags for the model; • conducting the Johansen cointegration test; • making a vector autoregressive (VAR) model and Granger test.
The time-series stationarity is one of the main conditions for constructing a vector autoregres-sive model. The Dickey-Fuller test checks its presence as it involves unit root identification. Using the STATA/IC 12 software, the following values of this test are obtained for the country sample variables (Table 9).
All criteria in the first step show non-stationarity of the time series data and need to be adjusted for further analysis using the first differences method. After that, all-time series are recognized as stationary and can be used in the following stages of research.
Determining the optimal number of lags is an important step in VAR analysis because it influences the model and its parameters. For their optimal selection, it is essential to analyze the level of significance (p) and information criteria for each model: Final Prediction Error (FPE), Akaike's information criterion (AIC), Hannan -Quinn information criterion (HQIC), and Schwarz Bayesian information criterion (SBIC). Table 10 shows an example of choosing the optimal number of lags for France for the three models. The asterisks indicate the series with the most optimal lags that have a significant value of p-statistics and the lowest values of information criteria. This algorithm is similar for all other countries.
Accordingly, the optimal number of lags for the three models varies from one to eight. Table 11 presents the results of choosing the optimal number of lags for the country sample obtained from the STATA software given in the appendices. The choice of the optimal number of lags is made by assessing the quality of the VAR model. The next step is a Johansen cointegration test or trace test, which allows the analysis of the longterm equilibrium between variables relationship. If it is absent, there is a need for further VAR modelling, which is verified by comparing the value of trace and the maximum eigenvalue statistic (max) with critical values. The results of this test are presented in Table 12.
All the values are below the critical values, so there is no cointegration. All this allows moving directly to VAR analysis. Eventually, significant results are not found for all countries considering the optimal lags. Table 13 shows the results that allow identifying the type of relationship (Y → X or X → Y), its character (direct or indirect) and the lag on which this trend appears.
As a result, it is confirmed that due to the investment position change per unit, the ESG index for Sweden decreases by 1.13 times with a lag of one year; for

DISCUSSIONS
Large-scale investment support by European countries in response to the pandemic is aimed at improving the EU's net international invest-ment position through multiplier effects and intensification of the transition to sustainable development. An essential issue in this context is the study of the relationship between traditional and ESG indices as the identification of a country's investment activity and its net investment position.
A bibliometric analysis of 1,747 publications within the topic of net international investment position and behavior of traditional and responsible indices over the period 2000-2021 by In-built Scopus instruments by Elsevier, In-built Web of Science instruments by Clarivate Analytics, Publish or Рerish, Google Scholar, VosViever shows that the study of this aspect is not present in the academic literature.
Existing studies do not provide evidences in favor of a direct positive impact of responsible in-