Foreign Investment, COVID-19 Stringency Measures and Risk of Openness

This paper analyzes the impact of the COVID-19 government stringency index on foreign investment, at the onset of the pandemic. Through a Robust Least Square regression estimation, we highlight that the relationship between foreign investment and the government containment measures displays a significant cross-country heterogeneity that depends on the level of the pandemic risk in the country. Foreign investors have indeed tilted their asset allocation towards those countries that implemented strong containment measures (as measured by the government stringency index) in the presence of a high risk (as measured by the risk of openness index). Conversely, they have shown a lower propensity to invest in assets issued by countries either adopting weak stringency measures despite a high risk of openness, or implementing drastic stringency measures in the presence of a relatively lower risk of openness. The above findings suggest the following interpretation: the government stringency measures and the pandemic risk have jointly affected foreign investors‟ behavior, which appeared relatively more prone towards assets issued by those economies better calibrating the policy interventions according to the pandemic harshness.

represents a new and unprecedented source of investor risk for international enterprises and then has likely hit investor confidence (Saurav et al., 2020). OECD (2020a) pointed out that the COVID outbreak has brought erosion of confidence and a great deal of uncertainty in international investments. In particular, emerging economies have already witnessed a massive reduction of portfolio investment inflows, because the higher uncertainty induced foreign investors to move capital back home, or invest in safer assets (OECD (2020b) and OECD (2020c)). Giofré (2022) confirms that, within a generalized decline in foreign investment, advanced countries, with higher GDP per capita, members of the G7 group, or of the Euro area have been significantly less severely hit by the pandemic than emerging and developing countries.
Moreover, OECD (2020b) and OECD (2020c) anticipated a sharper decline of foreign direct investments because of the stringent public health measures to limit the spread of the COVID-19, with a remarkable degree of heterogeneity across countries. Indeed, the Coronavirus outbreak has forced many governments to impose restrictions with different intensity and timing . Kizys et al. (2021) specifically investigate the impact of the COVID-19 governments" stringency measures on international stock markets and unveil that, in the first quarter of 2020, more severe containment policies have succeeded in reducing the multidimensional uncertainty, thus mitigating the investors" herding behavior. Giofré (2021) finds that inward investments have not been affected by the average stringency measures adopted by the issuing country. However, the within-country standard deviation of the stringency index has displayed a positive and significant correlation with a particular subset of foreign investments, that is, foreign portfolio investments. On the one hand, this stronger responsiveness is consistent with the literature that stressed that foreign portfolio investments are more reactive and volatile than foreign direct ones. On the other hand, their marked sensitivity to the within-country standard deviation of the stringency index suggests that foreign portfolio investors prefer governments" prompt reactions (characterized by large standard deviations) than gradual ones (characterized by lower standard deviations): a plausible interpretation is that the former policies are perceived as more effective to contain the uncertainty induced by the virus"s spread.
This paper contributes to the literature by investigating how an epidemic-specific source of risk, that is, the risk of openness, has affected the relationship between foreign investment and the government containment measures. We disclose that the non-significant correlation between inward foreign investment and the stringency index found by the literature (Giofré , 2021) hides an important source of heterogeneity across countries, which is strictly connected with the pandemic-induced uncertainty and risk aversion. The response of investors" behavior to the adoption of severe measures was related to the level of pandemic risk in the economy. Specifically, the extent to which severe containment measures -as measured by government stringency measures (SI)significantly affected the inward foreign investments in a country crucially rested on the country level of risk directly connected with the non-adoption or removal of these stringency measures -as proxied by the risk of openness index (ROI).
At the onset of the epidemic, foreign investors have highly valued the assets issued by countries, which calibrated the stringency measures according to the risk of openness. On the one hand, they highly rated the implementation of strong containment measures in countries with high risk of openness; on the other hand, they appeared to avoid investing in those countries adopting weak stringency measures despite a high risk of openness, or those implementing drastic stringency measures, in the presence of a low risk of openness.
The rest of the paper is structured as follows. Section 2 defines the dependent variable and describes the regression model. Section 3 reports the data and some meaningful descriptive statistics. Section 4 analyzes the empirical findings and provides robustness checks. Section 5 summarizes and concludes.

Regression Model
We define the growth of liabilities (ΔL) in a given period as the change in liabilities divided by the beginning of period liabilities" position. When considering the first quarter (q1) of 2020, therefore, we construct Δq1 as the difference between the liabilities" position at the end of the first quarter (March 2020) and the liabilities" position at the end of 2019 (December 2019), scaled by the liabilities" position at the end of 2019: Since foreign investments can be influenced by seasonality, we define, as an alternative to ΔL, diffΔL, that is, the difference between the ΔL measures in 2020 and 2019, as defined in equation (1) or (2), for the quarterly or semi-annual horizon, respectively. Therefore, diffΔq1, for the first quarter, is defined as follows: while diffΔs1, for the first semester, is defined as: The dependent variable, in its different definitions, is regressed on several explanatory variables. The main regressors in this analysis are the average within-country stringency index (SI), the risk of openness index (ROI), and their interaction (SI· ROI). Their coefficients are estimated through the following model: where  represents the error term of the regression estimation and controls capture a bunch of controlling covariates added to the econometric specification.
Ordinary Least Squares estimators are sensitive to the presence of outliers. We estimate the parameters in equation (5) through a Robust Least Squares estimation, a regression method specifically designed to be robust to the presence of outliers. Among the various Robust Least Squares models, we adopt the M-estimation (Huber, 1973). 1 For comparison with our baseline findings, we report the estimates under alternative estimation methods, such as OLS and Quantile regressions.
We are mainly interested in the sign, significance and size of the , γ and δ coefficients.  and γ capture the effect of the Stringency Index (SI) and of the Risk of Openness Index (ROI) on inward investments. Moreover, if the risk of openness in one country affects the way stringency measures influence foreign investments, then we should observe a significant δ, the coefficient of the interaction term SI·ROI.
We can include a limited number of controls because of the low number of observations, but the construction of the dependent variable, in difference form, partials out all country-specific fixed effects, as these are removed by construction.
Among the controls reported in equation (5), we include the Nominal Effective Exchange Rate (NEER) appreciation, the number of new COVID-deaths (and its within-country standard deviation), and two binary indicators of economic and financial development, to control for the presence of any eventual flight to quality inclination by foreign investors (Giofré , 2022).

Data and Descriptive Statistics
Our country sample includes 53 countries. We study the change in their foreign inward investments, which reflect their foreign liabilities" position, at the onset of the COVID crisis. The International Investment Position Statistics, released by the IMF, provides quarterly data on foreign assets and liabilities, by categories and instruments. For most of the analysis, we consider the wider Foreign Total Liabilities (FTL) asset class, but, in the last table, we also consider its sub-components, i.e., Foreign Direct (FDL) and Foreign Portfolio Liabilities (FPL).
The source of COVID-related data is the Coronavirus Open Citations Dataset, a Github ongoing repository of data on coronavirus. The Stringency Index and the Risk of Openness Index, the two main regressors of our analysis, are drawn from this source. They rely on the Oxford"s Coronavirus Government Response Tracker (OxCGRT) (Hale et al., 2020a: SI captures the severity of the government containment policy measures, while ROI is based on the recommendations set out by the World Health Organization (WHO) of the measures that should be put in place before Covid-19 response policies can be safely relaxed. From the same source, we also draw the epidemic data about new COVID-deaths and cases per million of inhabitants. While these are originally collected at a daily frequency, we consider their corresponding quarterly averages, for consistency with the quarterly dependent variable.
As anticipated above, we include as further controls, the NEER (Nominal Effective Exchange Rate), released by the Bank for International Settlements, and two binary indicators of economic and financial development, i.e., the GDP per capita and the market capitalization per GDP, both drawn from CEIC data. Giofré (2021), adopting the same dataset, provides a graphical evidence of the main statistical characteristics of foreign portfolio liabilities and of the stringency index. In Figure 1, we report the distribution and the basic descriptive statistics of the newly introduced explanatory variable, that is, the average quarterly Risk of Openness (ROI), in the first and second quarter. We observe that, while the mean of ROI slightly increases in the second quarter of 2020, the median in almost unaffected, and the standard deviation decreases.
In Table 1, we report the correlation matrix of the COVID-related regressors included in the analysis. Statistically significant correlation coefficients are displayed in bold characters (with p-values in brackets). The correlation between new COVID-deaths and new COVID-cases per millions are significantly correlated in both quarters (0.578 and 0.296). SI is significantly correlated with new COVID-deaths, only in the first quarter (0.219), and is never significantly correlated with ROI. The latter is significantly correlated with new COVID-cases per million in both quarters (0.240 and 0.584), and with new COVID-deaths, only in the second quarter (0.361). 2 Table 1

. Correlation matrix of COVID regressors
Note. This table reports the correlation matrix of COVID-related regressors. The upper panel refers to the first quarter of 2020, while the second one refers to the second quarter. Statistically significant Pearson-correlation coefficients are reported in bold characters (t-test p-values, in square brackets). Table 2 reports the main results of a multivariate regression analysis run under a Robust Least Squares estimation model. The dependent variable is the growth in Foreign Total Liabilities (FTL), as defined in equation (1), at the end of the first quarter (Δq1). As anticipated in Section 2, the definition of the dependent variable in difference form ensures that all country-specific fixed effects are removed by construction.

Main Results
The first main regressor included is the Stringency Index (SI), based on the Oxford"s Coronavirus Government Response Tracker, which is the quarterly average of daily data. We include, as a first control, the appreciation of the economy"s currency against a broad basket of currencies, as captured by the (one-month lagged) growth in the Nominal Effective Exchange Rate (NEER), because it may have affected foreign investments. 3 Second, we control for the quarterly average of new COVID-deaths per million of inhabitants. 4 Indeed, the recent literature on the effect of the COVID outbreak on financial markets highlighted a significant impact of COVID confirmed cases or deaths on volatility and liquidity (Ashraf, 2020;Albulescu, 2021;Baig et al., 2021;Salisu & Vo, 2020). As shown by the correlation matrix in Table 1, SI is correlated with these pandemic indicators, as it represents the measures adopted by governments as a reaction to new cases, intensive-care treatments, and deaths. Finally, since country-specific factors are swept away by construction of the dependent variable, rather than including the level of development of individual countries, we consider two binary indicators of development, GDP per capita and market capitalization to GDP. These are set equal to one, if the country belongs to the developed group, and 0 otherwise. If the flight to quality hypothesis is confirmed, as suggested by Giofré (2022), in the presence of a global shock, we expect foreign investors to deviate their investments to more stable and developed economies. 5 We can notice that, in column (1) of Table 2, the coefficient of the stringency index is not significant, and the only significant coefficients are the ones related to economic and financial development. In column (2) of Table  2, we add the quarterly average Risk of Openness Index (ROI) as a covariate, in order to check whether the risk of relaxing containment measures and opening to social and economic activities has somehow affected the growth in foreign liabilities. We observe that both indexes are not significant determinants of the growth in total foreign liabilities.
We conjecture however that the non-significant effect of the stringency measures on foreign liabilities could hide a source of cross-country heterogeneity, which depends precisely on the risk of openness faced by the country. Investors aim at reducing the exposition to risk, so that their behavior can be particularly reactive to government actions aimed at challenging the severe sources of pandemic risk. The relevance of strong containment measures as measured by SI -for investors could therefore crucially depend on the level of risk directly connected with the non-adoption or removal of these measures -as measured by ROI. Note. This table reports the results of a RLS regression (M-estimation), following equation (5). The dependent variable is the quarterly growth in Foreign Total Liabilities, defined as in equation (1). In columns (5a) and (5b), the Stringency Index (SI) and the Risk of Openness Index (ROI) are re-decoded as binary variables 0-1, if, respectively, their average quarterly value is below or above the mean (column (5a)), or median (column (5b)). ***, **, and * indicate significance at the 1, 5, and 10% levels, respectively.
To test this hypothesis, we add the interaction term between the stringency index SI and the risk of openness index ROI. If our conjecture is correct, we should observe a positive significant coefficient of the interaction term, thus suggesting that foreign investors consider the impact of the containment measures effective (or, at least, more effective) in economies experiencing a higher risk of openness.
While the results in column (3) do not fully support our initial conjecture, they suggest that it might be a promising direction. The coefficient of ROI is negative and significant (-0.383), while the coefficients of SI and of the interaction term SI·ROI are not statistically significant. 6 However, the non-significant coefficients are close to statistical significance, and their sign is consistent with our predictions: the coefficient of SI is negative, with a p-value equal to 0.121, while the coefficient of SI·ROI is positive and its p-value is 0.104.
Since Ashraf (2020) finds a difference between the stock market reaction to the growth in number of confirmed cases and its reaction to the growth in number of COVID deaths, in column (4), we replace the number of "new COVID-deaths per million of inhabitants" with the covariate "new COVID-cases per mn of inhabitants". We find that our coefficients are qualitatively unaffected by the inclusion of this alternative covariate, in terms of coefficients" significance and size.
It is important to notice that, the data source reports an explanatory note of the ROI index, which we directly quote here: it specifies that ‹‹the OxCGRT data cannot say precisely the risk faced by each country, it does provide for a rough comparison across nations. Even the "high level" view reveals that many countries are still facing considerable risks as they ease the stringency of policies›› (Hale et al, 2020a). We check therefore if, by accounting for the unavoidable "measurement" error implicit in the construction of such an index, our hypothesis finds further support. We construct therefore a binary variable for each of the two indexes, ROI and SI, splitting the countries into those above and those below the mean.
Interestingly, we observe in column (5a) that, when both indexes are dichotomized, the results corroborate our hypothesis. To interpret the effects, it is worth considering that the default-excluded group in the regression is the subset of countries with both SI and ROI below the mean. We observe, first, that, in the first quarter of 2020, countries with a high SI and a low ROI feature a 5.79% lower growth in FTL, while countries with high ROI and low SI display a 4.3% lower growth in FTL. Those countries adopting high SI in the presence of high ROI, instead, display a significantly larger growth in inward foreign investment, as shown by the coefficient of the interacted term (10.54%). To seize the overall growth in FTL, for countries with high SI and high ROI relative to countries with high SI and low ROI, we need to add the coefficient of the interaction term to the "pure" effect of the SI index. When compared to countries with high SI and low ROI, which witness a decrease in FTL by 5.79%, those countries featuring high SI and high ROI display a 4.75% larger FTL (= -5.79%+10.54%). This finding supports our conjecture that foreign investors are affected by the implementation of stringency indexes, to the extent that these measures are meant to reduce a high risk of openness. Conversely, foreign investors in general appear to discard the assets issued by those countries whose containment policies are relatively mismatched with the risk of openness, that is, those adopting weak stringency measures despite a high risk of openness, or those implementing drastic stringency measures in the presence of a lower than average risk of openness.
In column (5b), we check whether our findings are sensitive to a different specification of the threshold to define the binary version of the two indexes. If we consider the median, as an alternative to the mean, we observe that the findings are substantially unchanged.
Finally, columns (6a) and (6b) replicate the regressions of columns (5a) and (5b), when the "number of new COVID-cases per million" replaces the "number of new COVID-deaths per million", and results are confirmed, with only modest changes in the size of the coefficients. Note. This table is the same as Table 2, but the dependent variable is defined as diffΔq1, as defined in equation (3). Table 3 replicates Table 2, but the dependent variable is defined as diffΔq1 (equation (3)), rather than as Δq1 (equation (1)). This measure addresses the issue of the seasonality of foreign investments, as it is derived as the difference between the 2020 first quarter measure, and the corresponding first quarter measure in 2019. This table provides results very similar to the ones reported in Table 2. In the first two columns, we observe no significant coefficients, but in the third column, we find some hints in support of our hypothesis, as the coefficients of the SI and of the ROI index are negative, while the coefficient of the interaction term SI· ROI is positive. However, when replacing the "new COVID-deaths" control with the "new COVID-cases", all coefficients, with the exception of ROI"s, fall below the standard bar of statistical significance. As in Table 2, when considering the indexes in a dichotomic version, the sign and significance of coefficients are restored and become fully consistent with our conjecture. In column (5a), we observe that, in the first quarter of 2020, countries with high SI and low ROI witness a decrease in FTL by 6.41% with respect to 2019, while those countries with high SI and high ROI indeed display an increase in FTL by 2.09% (= -6.41%+8.50%).
A comparison with Table 2 reveals that the effect is still present and statistically significant, thus supporting our hypothesis, although the effect on the growth of FTL in the first quarter of 2020, is halved in size. Results are confirmed when considering the median threshold, rather than mean (column (5b)), or when considering the alternative covariate "new COVID-cases", under both specification of the threshold (columns (6a) and (6b)). Table 4 replicates the analysis of Table 2 and 3, when the dependent variable is the growth in liabilities at a one-semester time span. For the sake of brevity, we report only the relevant regressors. The upper part of the table (panel I) refers to the Δs1 measure (equation (2)), while the bottom part (panel II) refers to the diffΔs1 measure (equation (4)), defined as the difference between the growth of FTL in the first semester 2020 with respect to the first semester in 2019. We observe that, differently from the first quarter, the SI and ROI indexes have no significant role in driving foreign investors" decisions at the end of the first semester, under any specification of the indexes, or of the dependent variable. 7 Table 4. Main findings: Δs1 and diffΔs1 Note. This table replicates Table 2 and 3, but relative to the first semester of 2020. In panel I, the dependent variable follows equation (2), while in panel II, it follows equation (4). Controls reported at the bottom of the table are included, but not reported.

Robustness Checks and Additional Analysis
The empirical evidence above has shown that the measures of containment (SI) and the risk of openness (ROI) may help explain the foreign investors" choice, at the onset of the COVID outbreak, but not in the second quarter. 8 We now subject our findings to a bunch of robustness checks and additional studies, to understand the strengths and limits of the analysis.
In Table 5, we undergo the sensitivity of the (significant) findings for the first quarter to different estimation strategies and country-sample specification. As in Table 4, the upper part of the table (panel I) refers to the Δ measure, while the bottom part (panel II) refers to the diffΔ measure. The first three columns consider alternative estimation models, while columns (4) to (5c) consider different country-sample specifications. At the head of the rows, we specify that the indexes are defined in a dichotomic form (SI_d and ROI_d). In column (1) of panel I and II, we report, for comparability, the results of column (5a) of Table 2 and Table 3, following the Robust Least Squares (RLS) baseline approach. Column (2) reports the results of the regression under an OLS specification, which are qualitatively similar to the ones in column (1). Column (3) reports the results under a Quantile regression (computed at the median). We show that, in the first panel, the results relative to the median of the response variable are fully in line with our previous findings, both in terms of significance and size. Conversely, panel II, in which the dependent variable is defined as the difference between the growth of FTL in the first quarter of 2020 and the corresponding growth in 2019, displays less robust results, with a (marginally) significant coefficient of the interaction term SI·ROI, and a (marginally) non-significant coefficient of the SI term (p-value 0.11).
In columns (4) to (5c) of panel I and II, we test whether our findings are robust to the country sample specification. In column (4), we exclude China, which has been the first economy hit by the COVID outbreak, well before other countries. China"s asynchronous timing of lockdown and loosening measures might have distorted or driven our results. We observe instead that the exclusion of China, in both panels, hardly affects the size and significance of the coefficients. Note. This table reports the results of the sensitivity analysis to different econometric and sample specifications. For the sake of brevity, only results with a binary definition of SI_d and ROI_d (1 if above the mean, 0 otherwise) are reported.
In columns (5a) to (5c) of Table 5, we exclude from the sample potential offshore financial centers, according to three different classifications. Column (5a) reports the results under the offshore classification proposed by Damgaard et al. (2018), column (5b) follows Zoromé (2007), while column (5c) follows Lane & Milesi-Ferretti (2017). 9 By comparison with column (1), we observe that the results are confirmed, and interestingly, the exclusion of offshore centers has even reinforced them: both the negative coefficient of the SI index and the positive coefficient of the interaction term SI·ROI are economically larger and statistically more significant. Giofré (2021) finds that, at the end of the first quarter of 2020, the standard deviation of the stringency index (σSI) is positively and significantly correlated with a subset of foreign inward investment, that is, portfolio investments. She suggests that foreign portfolio investors, typically more reactive than foreign direct investors, could have been more responsive to governments" prompt reactions (featuring a larger SI standard deviation) than to gradual ones (featuring a smaller SI standard deviation), because the former policies may have been perceived as a more serious commitment to limit the uncertainty due to the spread of the virus.
In Table 6, we check if our results are affected by the inclusion of the standard deviation of the stringency index (σSI). If countries with a higher SI standard deviation are also those with a higher risk of openness, the omission one of the two factors can make the coefficient of the included covariate biased. 10 We consider therefore a regression specification including both regressors, to test if the two pieces of empirical evidence ̶ the one reported in Giofré (2021) and the one described in this paper ̶ can coexist or are mutually exclusive. Since Giofré (2021) underlines a different role for portfolio and direct inward investments, we partition Table 6 horizontally in three panels, and report the results for foreign total liabilities (FTL, panel I), foreign direct liabilities (FDL, panel II), and foreign portfolio liabilities (FPL, panel III).
The evidence on the coefficient of the σSI regressor is fully in line with the results of Giofré (2021) for FTL, in terms of sign (positive), significance (marginal), and size (about 0.3%) of the associated coefficient. In panel II and III, we compare Foreign Direct Liabilities and Foreign Portfolio Liabilities. Giofré (2021) finds that foreign direct inward investment have shown a lower responsiveness to σSI than foreign portfolio inward investments. In our analysis, we confirm that the coefficient of the σSI is almost twice as large for FPL than for FDL, and its statistical significance is much stronger and systematic.
We observe in Table 6, after the inclusion of the additional covariate σSI, the joint role of SI and ROI, which is the focus of the paper. In panel I, the regression setting is in fact the same as in Table 2, with the addition of the standard deviation of the stringency index (σSI) as a regressor. We observe that the coefficient of the SI regressor in the first two columns is not significant, as in our previous findings. Also after including the interaction term between SI and ROI, we do not observe any significant coefficient of SI, either when controlling for the new COVID-deaths (column (3)), or when controlling for the new COVID-cases (column (4)). Again, when we re-code the SI and ROI indexes in a dichotomic form (columns (5a) to (6b)), we find a statistically significant coefficient of the interaction term SI·ROI, consistently with our hypothesis, while the negative coefficient of the SI is statistically different from zero only if the mean is used as a threshold of the binary variables (columns (#a)). When considering the dichotomic version of the SI and ROI indexes in columns (5a) to (6b), we observe the following differences between FDL and FPL. In terms of significance, the coefficient of the interaction term SI·ROI for FDL is significant only in columns (#a), while it is always statistically different from zero for FPL (though marginally, when the threshold is the median). In terms of size, the overall effect of countries with high SI and high ROI is larger for FPL than for FDL. Indeed, although the negative coefficients of the SI regressor and the positive coefficients of the interaction term SI·ROI are larger (in absolute value) for FDL than for FPL, the overall net effect of high SI and high ROI on foreign liabilities" growth is smaller for FDL than for FPL. The difference is only marginal in column (5a), with a 0.0210 (= -0.0819+0.1028) for FDL versus a 0.0267 (= -0.0639+0.0906) for FPL, while it is more evident in column (6a), with a 0.0103 (= -0.0899+0.1003) for FDL versus a 0.0225 (= -0.0554+0.0780) for FPL.
Moreover, the growth in foreign portfolio liabilities in the first quarter of 2020 is significantly associated with the stringency index SI, the risk of openness ROI, and their interaction, even in their continuous version. In columns (3) and (4) of panel III (FPL), the coefficients of SI, ROI, and SI·ROI, defined in their original continuous form, are all statistically significant, differently from FDL (panel II) and FTL (panel I), and, more generally, to the results in the whole analysis conducted so far. It means that, while the growth in foreign direct liabilities only responds to high versus low indexes, foreign portfolio liabilities are tilted also by a marginal difference of the stringency index, across economies differing by a marginal degree of openness risk exposure.
For instance, the results in column (4) of panel III, can be read as follows: an increase in the SI index, originally ranging from 0 to 100, by 1 unit leads to a -1.46% lower growth in FPL, if the risk of openness ROI is set at its minimum (that is equal to 0); the same unit increase in the SI index leads to an increase in the growth of FPL by 0.7% (0.0070= -0.0146+0.0217), when the level of risk of openness is set at its maximum (that is equal to 1).
This evidence points to a tighter and closer responsiveness of foreign portfolio liabilities rather than direct portfolio liabilities to the stringency measures adopted and to the COVID risk exposure of the country, consistently with the results in Giofré (2021), thus confirming a general higher reactivity by foreign portfolio investors to the COVID outbreak.
Overall, Table 6 shows that our findings and the evidence in Giofré (2021) are mutually consistent. On the one hand, the role of the standard deviation of the stringency index σSI is confirmed as a significant driver of foreign portfolio investment, even after considering the risk of openness. On the other hand, the inclusion of σSI does not invalidate our findings, but, on the contrary, enriches the analysis, by unfolding the multifaceted sensitivity of foreign portfolio investors to the adoption of COVID containment measures, at the onset of the virus spread.

Conclusions
We find that the response of investors" behavior to the adoption of COVID restrictive policies crucially depends on the level of the pandemic risk of the economy. Specifically, the extent to which severe containment measures (as measured by the government stringency index, SI) significantly affect inward foreign investments in a country depends on its level of pandemic risk, which is connected with the non-adoption or removal of these stringency measures (as proxied by the risk of openness index, ROI).
Foreign investors prefer countries that calibrate the stringency measures according to the risk of openness. On the one hand, they highly rate the implementation of strong containment measures in countries with high risk of openness. On the other hand, they appear to avoid investing in those countries either adopting weak stringency measures despite a high risk of openness, or implementing drastic stringency measures in the presence of a low risk of openness.
The limited time span, the cross-sectional nature of the analysis the low frequency of international asset holdings" data, on the one hand, prevents us from adopting more sophisticated statistical and econometric tools, and on the other hand, it may affect the general validity and policy implications of our findings. The availability of a longer time span may stimulate further research, and provide a wider perspective on the main drivers of the evolution of international investment during and after the COVID pandemic.
Notwithstanding the above-cited limitations, this study may provide some insights on the immediate reaction of foreign investors to the pandemic outbreak. The severity and speed of adoption of policies is strictly connected with the severity of the effects of the COVID spread, whose remarkable heterogeneity across countries has not been fully understood, yet. This paper emphasizes the importance of taking into account this multifaceted heterogeneity, by considering how the diversity in the risk of openness across countries may have affected the linkages between government containment policies and foreign investment decisions.