Financial-integration thresholds for consumption risk-sharing

REVECO-01034; No of Pages 21 International Review of Economics and Finance xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

International Review of Economics and Finance journal homepage: www.elsevier.com/locate/iref

Financial-integration thresholds for consumption risk-sharing Samreen Malik ⁎ New York University-Abu Dhabi, PO Box 903, New York, NY 10276, USA

a r t i c l e

i n f o

Article history: Received 6 October 2013 Received in revised form 16 January 2015 Accepted 20 January 2015 Available online xxxx JEL codes: F3 F4 F6

a b s t r a c t I present empirical evidence of how international consumption risk sharing varies by levels of financial integration. In a panel data set of 64 countries from 1985–2009, I show a significant presence of threshold effects of financial integration on international consumption risk sharing. The results indicate the presence of two significant thresholds and three corresponding regimes. Below the lower threshold is limited but statistically significant consumption risk-sharing. Above the higher threshold is significant risk-sharing. However, intermediate to the two thresholds is a regime with excess volatility. These findings are therefore suggestive of a U-shaped relationship between financial integration and consumption risk-sharing, with a potentially destabilizing intermediate regime. © 2015 Elsevier Inc. All rights reserved.

Keywords: Financial integration Consumption risk-sharing Threshold effects

1. Introduction Over the past three decades, a large increase has occurred in the stock of cross-border financial capital holdings, especially in emerging and developing economies. In theory, such financial integration should promote international consumption risk-sharing. Cross-border financial linkages can decouple domestic consumption from country-specific components of output shocks, thereby decreasing the correlation between domestic consumption and output, and increasing the correlation between domestic consumption and world consumption. However, the empirical evidence on the relationship between financial integration and international consumption risk-sharing is mixed. Existing research suggests that risk-sharing has not increased much, especially in emerging economies, despite the wave of financial globalization. In fact, some evidence shows greater financial integration can also increase the correlation between domestic consumption and domestic output, and can even lead to excess volatility (i.e., reduced consumption risk-sharing).1 The puzzling empirical relationship between financial integration and international consumption risk-sharing has led some people to posit the existence of a threshold effect, whereby consumption risk-sharing is achieved only after a certain degree of financial integration is reached (see for e.g., Kose, Prasad, & Terrones, 2007, 2009b). In this paper, I formally test for the existence of such threshold effects in a panel data set of 64 countries from 1985–2009. In particular, I estimate thresholds by adapting a method for threshold estimation developed by Hansen (1999) for balanced panels. This methodology allows for estimation of thresholds, confidence intervals, and standard errors directly from the data, and thereby eschews an ad-hoc categorization of countries based on GDPper-capita (versus the categorization based on actual measures of financial integration of interest).

⁎ Tel.: +971 50 441 5197. E-mail address: [email protected]. 1 See, for example, recent work by Kose, Prasad, & Terrones (2003, 2007, 2009b) and Bai & Zhang (2012).

http://dx.doi.org/10.1016/j.iref.2015.01.004 1059-0560/© 2015 Elsevier Inc. All rights reserved.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

2

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

To implement the methodology empirically, I proxy financial integration by various quantity-basedfinancial-integration measures,2 and measure consumption risk-sharing in terms of the coefficient on the annual country-specific consumption growth rates against country-specific output growth rates in a panel controlling for individual (and time) fixed effects (Demyanyk, Ostergaard, & Sorensen, 2008; Kose et al., 2007). I find that economies with highly integrated capital markets have fostered imperfect but better risk-sharing than countries with less integrated capital markets. More formally, the results show that financial integration when measured using either (1) the sum of gross total foreign direct investment (FDI) and foreign portfolio investment (FPI) (as a ratio of GDP) (henceforth equity volume) or (2) total foreign assets and liabilities (to GDP) (henceforth total volume) reveal two significant thresholds corresponding to three regimes. A lower threshold value of financial integration characterizes the first regime (e.g., the estimated threshold level in the case of equity volume to GDP as the financial integration measure is approximately 43.5%). Country–year pairs falling in this regime show statistically significant and positive but economically small international consumption risk-sharing. In the third regime (e.g., the estimated level of threshold for equity volume to GDP is approximately 84.1%), country–year pairs' risk-sharing remains imperfect but is both statistically and economically significant, with an average consumption risk-sharing coefficient of 50% (coefficient of 0 corresponds to perfect consumption risk-sharing, whereas a coefficient of 1 indicates no consumption risk-sharing). However, intermediate to the two thresholds is a regime with excess volatility (relative to Regime 1 and Regime 3), where average correlation between country-specific consumption growth rates and output growth rates exceedcountry-specific consumption and output growth rates in Regime 1 and Regime 3. These findings therefore suggest a U-shaped relationship between financial integration and consumption risk-sharing, with a potentially destabilizing intermediate regime that prior literature has missed. Consumption risk-sharing estimated in Regime 1 and Regime 3 is in line with the theoretical predictions that better financial integration fosters better consumption risk-sharing. However, these predictions are at odds with Regime 2, which has better financial integration but worse consumption risk-sharing relative to Regime 1. Further investigation into the composition of capital stock across these regimes reveals that the key difference between these three regimes is the quantity of FDI liabilities (as a percentage of total assets and liabilities). In particular, on average Regime 2 have the highest FDI liabilities. This regime can be reconciled with the recent theoretical predictions that initial capital in the form of FDI liabilities can in fact reduce consumption risk-sharing (see e.g., Acemoglu & Zilibotti, 1997; Milesi-Ferretti & Razin, 1998; Razin & Sadka, 2001). The contributions of this paper are twofold: First, the finding of significant threshold effects suggests a nuanced approach to interpreting the empirical relationship between financial integration and consumption risk-sharing more generally than has been done in the existing literature. In the existing literature, patterns of consumption risk-sharing are generally studied using simple, linear interaction of idiosyncratic components of output with measures of financial integration, or using separate estimates for subsamples of developing, emerging, or industrial economies (see e.g., Kose et al., 2007, 2009b). Threshold effects, however, indicate the presence of statistically significant non-linearity in the relationship between financial integration and risk-sharing. Such nonlinearities are consistent with the mixed findings in the existing literature, but suggest an alternative rationale whereby economies face an intermediate hurdle on the path of financial integration before benefits of consumption risk-sharing are realized. Second, unlike existing literature (Kose et al., 2007, 2009b), this paper highlights that in addition to debt liabilities, primitive capital in the form of FDI liabilities may also be an additional source of worsening consumption risk-sharing, which is captured through the intermediate regime. The remainder of the paper is structured as follows. Section 2 discusses related literature. Section 3 outlines the data set, and Section 4 provides details on the empirical methodology. Section 5 presents the results. Section 7 concludes. Appendix A and B contain supporting tables and illustrations, respectively.3 2. Related literature A large literature exists on consumption risk-sharing in industrialized economies. The consensus from this literature is that risksharing in industrialized economies is imperfect (Ambler, Cardia, & Zimmermann, 2004; Backus, Kehoe, & Kydland, 1992; Canova & Ravn, 1996; Pakko, 1998) but has improved significantly with greater integration of financial markets (Demyanyk et al., 2008; Lewis, 1996; Obstfeld, 1994; Pakko, 1998). Ambler et al. (2004),Backus et al. (1992),and Canova & Ravn (1996) all study crosscountry consumption correlations in industrialized countries and conclude that risk-sharing remains low relative to the predictions of a standard Arrow–Debreu contingent claims economy. More recently, researchers have focused on consumption risk-sharing in emerging economies. Kose et al. (2003, 2007, 2009b) and Bai & Zhang (2012) show that, relative to industrialized countries, consumption risk-sharing in emerging economies is (1) small and (2) appears much less responsive to an increase in financial market integration. Kose et al. (2003, 2007) employ a variety of measures of consumption risk-sharing for 72 economies for the sample period 1960–2004. They divide the full sample of these countries into three sub-groups: advanced countries (21), emerging markets (22), and other developing countries (33). They conclude that advanced economies have achieved more consumption risk-sharing during the sample period. Similarly, Bai & Zhang (2012) find that 2 Two types of measures of financial integration exist. The first type are de-jure measures that attempt to capture legal restrictions on cross-border capital flows (see Chinn & Ito, 2008; Edison, Klein, Ricci, & Slok, 2002; Edwards, 2001; Quinn, 1997). The second type is de-facto measures that are based on actual cross-border capital stocks, and include the positive sum of items in the capital account viewed relative to GDP. Quantity based measures provide a less volatile and a more appropriate measure of integration for studying risk-sharing phenomenon (see, e.g., Kose, Prasad, Rogoff, & Wei (2009a) for a discussion of the favorability of such measures in the context of consumption risk-sharing). 3 I also provide an additional section in the appendix, where I compare the results of the current paper and the existing literature to highlight that the true process of financial integration is not fully captured in the existing literature because of a priori assumption about the existence and level of threshold.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

3

the coefficient estimated by regressing domestic consumption growth on domestic output growth is lower and the coefficient of domestic consumption growth on world consumption growth is higher for the industrialized countries than for emerging economies, suggesting greater risk-sharing in the developed economies. Studies have further investigated this puzzling nonresponsiveness of consumption risk-sharing in emerging and developing economies during the financial integration process (see, e.g., Suzuki, 2014; Kose & Prasad, 2010). In particular, Suzuki (2014) points toward the differences in income processes as the underlying factor for the varying degree of consumption risk-sharing and smoothing among economies as a result of financial integration.4Kose & Prasad (2010), on the other hand, use the composition of financial assets and liabilities to study consumption risk-sharing patterns. Of these various types of assets and liabilities, Kose & Prasad (2010) find that external debt is the least conducive to risk-sharing, but the risk-sharing benefits of increased FDI and FPI holdings for emerging and developing economies also remain negligible.5 However, theoretical predictions such as, Acemoglu & Zilibotti (1997), Milesi-Ferretti & Razin (1998), and Razin & Sadka (2001), point out that initial FDI liabilities may also be a potential reason for excessive volatility in the process of financial integration. In particular, Acemoglu & Zilibotti (1997) model that in a setting where the financing of the start-up cost of a new project is covered by foreign investors at the initial stages integration may result into higher volatility in the economy. The rationale behind increasing volatility is that the foreign investors concentrate their capital in the sectors of the host economy that provide higher growth and risk-sharing opportunities relative to the investor's domicile economy. To avoid volatility, the host economy should make counter investments in other projects that are less correlated with the host economy. Failure to do so (perhaps due to capital scarcity) exposes the host economy to more volatility, as a result of initial increases in FDI liabilities, instead of reducing volatility. In addition, Razin & Sadka (2001)'s framework can be interpreted as a setting where the foreign investors seek investment and have an access to the skimming technology to find high-growth sectors that also provide diversification opportunities. The domestic investors can only access costly skimming technology, and therefore rely on signaling from foreign investors' investment decisions. Subsequently, domestic and foreign investors invest in the same projects, which results in over-investment in specific sectors. In addition, Milesi-Ferretti & Razin (1998) point out that financial transactions can sometimes accomplish a reversal of FDI liabilities quite easily as well. Therefore, the initial investments in the form of FDI liabilities might induce additional income volatility and also inhibit the degree of consumption risk-sharing. Whereas the empirical findings on consumption risk-sharing, especially in emerging and developing economies, may be viewed as casting capital market integration in a negative light, Kose et al. (2007) instead suggest these findings may indicate that the process of integration in these economies has not gone far enough to realize potential gains of consumption risk-sharing. In particular, Kose et al. (2007), use linear interaction terms (of financial integration measure and the variable of interest) and find that the benefits of financial integration in terms of international risk-sharing are accrued only after a certain threshold of financial openness is reached. Similarly, Kose, Prasad, & Taylor (2011) identify threshold effects of financial depth and institutional quality for the effects of capital market integration on output growth rates. They employ parametric and non-parametric approaches, and conclude that beyond the estimated threshold level, presumed risks of financial openness are also reduced and indirect benefits of consumption risk-sharing could potentially be realized. A synthesis of this empirical literature therefore already indicates the potential for threshold effects of capital market integration on welfare-relevant measures such as output growth and international consumption risk-sharing. However, all of the existing literature uses exogenous specifications of threshold effects to in turn demonstrate the existence of threshold effects. A primary contribution of the present paper is to employ a methodology where, threshold effects can be robustly identified without imposing prior assumptions on the existence or levels of thresholds. Instead of using arbitrary threshold levels as a sample-splitting criterion, I employ a systematic methodology developed by Hansen (1999) in order to simultaneously identify and estimate threshold effects, levels, confidence intervals, and coefficients directly from the data. In addition, the results in this paper identify the presence of an additional intermediate regime (which the existing literature, such as Kose et al. (2007, 2011), has missed). This regime enhances the understanding of the process of financial integration and its implications on consumption risk-sharing. In particular, this regime highlights that dominance of debt liabilities observed in Regime 1 (also highlighted by Kose & Prasad, 2010) is not sufficient to explain the lack of consumption risk-sharing observed in the data. An additional regime exists that experiences an initial increase in FDI liabilities and therefore encapsulates an additional cost of financial integration resulting into lack of consumption risk-sharing. This empirical finding of intermediate regime is consistent with the theoretical literature (such as Acemoglu & Zilibotti, 1997; Milesi-Ferretti & Razin, 1998; Razin & Sadka, 2001) and also has direct policy relevance. 3. Data I estimate threshold levels and coefficients of consumption risk-sharing from a balanced panel data set of developing, emerging, and advanced economies from 1985–2009. Per-capita real GDP and per-capita real (public plus private) consumption are taken from the World Bank's Development Indicators. Stocks of financial assets and liabilities (debt, foreign direct investments (FDI), and foreign portfolio investment (FPI)) are taken from the extended “External Wealth of Nations” data set by Lane & Milesi-Ferretti 4 In particular, Suzuki (2014) tests the joint hypothesis of rational expectations and permanent income hypothesis. In this paper, the author decomposes the GDP into permanent and transitory components and specifies respective processes of each component using a state-space/unobserved component model to derive permanent slope and level-based income shocks and transitory shocks. Furthermore, controlling for such a stochastic nature of income paths Suzuki (2014) reveal that rational expectations (RE) and the permanent income hypothesis (PIH) for transitory income are not rejected for OECD countries, and the predictions of RE/PIH of large increases in consumption in response to positive income growth shocks also hold in the data. 5 For a summary of predictions, methodologies, and results on consumption risk-sharing in industrialized and emerging economies, see, for example, Kose & Prasad (2010) or Islamaj (2008).

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

4

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

(2007). Proxies of financial-integration measures are constructed in line with the scale of international financial-integration measures provided in Lane & Milesi-Ferretti (2007) who use the actual cross-border capital stocks, and include the positive sum of items in the capital account (relative to GDP). These financial-integration measures are in terms of total capital stock volume and equity volume (as percentage of GDP) (also referred to as de facto measures).6 The data set has annual data over the period 1985–2009 for 21 developing, 22 emerging economies and 22 advanced economies. Sources of data and list of countries are provided in Tables A.2 and A.1, respectively, inAppendix A. Table A.3 gives summary statistics for the variables, including 10th, 25th, 50th (median), 75th and 90th percentiles for each type of capital stocks and per capital variables (average over all years) in the sample. The distinction between developing, emerging and advanced economies is based on IMF classifications in each time period. The list of countries matches closely to the countries studied in other empirical work, such as Kose, Prasad, & Terrones (2009b). Minor omissions are due to the requirement of a balanced panel (therefore, lack of data availability or missing values result into exclusion of Cote d'Ivoire, Haiti, Iran, Panama, Papua New Guinea and Togo from the list of developing countries.). 4. Methodology I first describe a specification for estimating international consumption risk-sharing based on Kalemli-Ozcan, Sorensen, & Yosha (2003) and Demyanyk et al. (2008) in Section 4.1. In Section 4.2, I describe a methodology for identifying and estimating threshold effects from a balanced panel, which is introduced in Hansen (1999). Finally, Section 4.3 outlines a synthesis of these two methodologies that can be employed to identify and estimate threshold effects of financial integration on international consumption risk-sharing. 4.1. International consumption risk-sharing A common approach to measure consumption risk-sharing in the literature is to estimate a coefficient for a regression of countryspecific growth rates on country-specific output (see, e.g., Bai & Zhang, 2012; Demyanyk et al., 2008; Kalemli-Ozcan et al., 2003; Kose et al., 2003, 2011). Let i denote the country index and let t denote the time index for a balanced panel of observations on I countries over T time periods. Per-capita consumption (public and private) is denoted by cit. To focus on country-specific growth rates and remove uninsurable aggregate shocks, denote aggregate consumption of the total sample in period t by ct := ∑icit. The country-specific consumption growth rate is then given by: ½logðcit Þ−logðciðt−1Þ Þ−½log ðct Þ−logðct−1 Þ :¼ Δlog ðcit Þ−Δlog ðct Þ:

ð1Þ

In a similar manner, country i's year t, per capita GDP is denoted by yit, and aggregate GDP of the total countries in the sample is denoted by yt := ∑iyit for all t ∈ T. The following expression then captures the country-specificGDPper-capita growth rate: ½logðyit Þ−logðyiðt−1Þ Þ−½logðyt Þ−logðyt−1 Þ :¼ Δlog ðyit Þ−Δlogðyt Þ:

ð2Þ

International consumption risk-sharing can be measured with the coefficient of a regression of country-specific consumption on country-specificGDP growth rates, in a basic regression controlling for country and time fixed effects: Δlog ðcit Þ−Δlog ðct Þ ¼ μ i þ βc ðΔlogðyit Þ−Δlogðyt ÞÞ þ ϵ it ;

ð3Þ

where μi, is a country-specific fixed effect and ϵit is an error term, assumed to be conditionally i.i.d normally distributed across (i, t) ∈ I × T. Perfect consumption risk-sharing should imply zero correlation between the country-specific component of consumption growth rate and GDP growth rate, and can therefore be tested with the the null hypothesis that βc = 0, whereas a complete absence of risk-sharing corresponds to the null βc = 1. Coefficients between 0 and 1 correspond to partial risk-sharing, whereas coefficients greater than 1 indicate excess volatility. Demyanyk et al. (2008) extend the benchmark specification outlined above to study whether consumption and income risk-sharing increase with greater financial integration. They estimate the effect financial integration has on consumption risk-sharing by including interaction terms with foreign financial assets and liabilities relative to GDP (see also Melitz & Zumer, 1999). The empirical specification is therefore as follows: Δlogðcit Þ−Δlog ðct Þ ¼ μ i þ κ ðΔlogðyit Þ−Δlogðyt ÞÞ þ ϵ it ;

ð4Þ

where they impose structure on the coefficient κ, allowing it to vary over time, countries, and through the sum of foreign assets and liabilities (relative to GDP): κ ¼ κ o þ κ 1 t þ κ 2 ð F it −F i Þ:

ð5Þ

6 While the former measure includes debt securities which may be driven by other special factors, has less relevance for the risk-sharing motivation, however the analysis in this paper employs both measures separately to study the potential threshold effects on consumption risk-sharing.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

5

In expression (5), Fit denotes the capital account entry for a generic class of foreign assets or liabilities in country i ∈ I at time t ∈ T, and includes variables as deviations from country-specific means FAi = ∑TFAit. The structure includes a time trend to guard against any trend in assets and liabilities so that the analysis does not capture the changing trend of risk-sharing that may result from other developments in national economies. The estimated value of (1 − κo) then captures the average amount of consumption risk-sharing within the group of countries; κ1 captures time trends and κ2 captures the effect of changes in foreign liability or asset position on con
sumption risk-sharing. 1−κ o −κ 1 t−κ 2 F it −F i measures the amount of consumption risk-sharing that country i obtains in period t. Similar to the above methodology, the analysis in this paper is also based on fixed-effects model, and uses the specification (4) but with threshold effects of liability and asset positions in lieu of the linear-interaction specification of foreign asset effects in expression (4). At a basic level, the specification used in this paper, simply involves placing a different structure on the coefficient κ in expression (4). Although mean, median, or quantile values for thresholds may often be a convenient approximation, such threshold levels impose ad hoc restrictions on the data, especially when the threshold variable is a decision variable for agents. I therefore follow an alternative method for identification of endogenous threshold effects, based on the methodology introduced in Hansen (1999). 4.2. Threshold effects in non-dynamic panels Hansen (1999) develops econometric techniques for estimating threshold effects in a balanced panel, where regression coefficients are not identical across all observations in the sample but differ depending on discrete classes into which they fall. Thresholds are identified via an iterative bootstrap method, which also constructs consistent confidence intervals for the threshold parameters, based on the asymptotic distribution of the error term. The statistical significance of threshold parameters is assessed on the basis of the bootstrap method. This method allows estimation of parameters from the sample, which also has clear and tractable relevance for policy-related issues. The basic regression equation of interest is of the following form: 0

0

yit ¼ μ i þ β1 xit Iðqit ≤ γÞ þ β2 xit Iðqit N γ Þ þ ϵ it

ð6Þ

where I(.) is the indicator function, the subscript i ∈ I indexes the individual, and the subscript t ∈ T indexes time. In the first iteration, the observations are divided into two regimes depending on the threshold variable, qit. These regimes can be identified based on the regression slopes, which are denoted by β1 and β2. The basic identification assumption is that xit and qit are not time-invariant.7 The second iteration, tries to further divide the sample, continuing iterations until no further thresholds are identified. The method tests for the existence of thresholds and further determines the number of thresholds that are estimated by least squares. The slope coefficients are estimated along with conventional OLS standard errors and white-corrected standard errors to test the significance of these slope coefficients. In the above specification, the analysis tests for zero, one, two, or three thresholds. The F-test statistics and the likelihood ratio along with bootstrap p values indicate significance of the threshold parameters. If the bootstrap p value is below the desired critical value, the null hypothesis of no threshold, one threshold, two threshold or three threshold is rejected. Furthermore, the asymptotic confidence intervals for the threshold are then used to ascertain the certainty about the nature of the division based on the threshold parameters. 4.3. Consumption risk-sharing and threshold effects In this section, I outline a synthesis of the consumption risk-sharing methodology described in Section 4.1 and threshold identification methodology described in Section 4.2, which represents the main empirical methodology that I employ to identify threshold effects on consumption risk-sharing. The threshold variables are de facto measures of financial integration, such as foreign asset and liability stocks (relative to GDP). The threshold regression specification follows: Δlogðcit Þ−Δlogðct Þ ¼ μ i þ βo t ðΔlog ðyit Þ−Δlogðyt ÞÞ 1 þ βc ðΔlogðyit Þ−Δlogðyt ÞÞð F it ≤ γ1 Þ 2 þ βc ðΔlogðyit Þ−Δlogðyt ÞÞðγ 1 b F it ≤ γ 2 Þ 3 þ βc ðΔlogðyit Þ−Δlogðyt ÞÞðγ 2 b F it ≤ γ3 Þ 4 þ βc ðΔlogðyit Þ−Δlogðyt ÞÞðγ 3 b F it Þ þ ϵ it :

ð7Þ

The basic specification simply places a different structure on κ in expression (4), based on threshold effects instead of linearinteraction effects. The specification also captures time trends as suggested by Demyanyk et al. (2008) to control for changing trends in risk-sharing that might be the result of any other developments in the national economies. The estimation of threshold effects and confidence regions is then based on Hansen's (1999) iterative bootstrap procedure, which produces robust and consistent estimates of threshold values, confidence intervals, and coefficients in each threshold regime. The statistic (1 − βc) is interpreted as the average amount of consumption risk-sharing within the regime, where respective β's differ on the basis of various regimes determined by the estimated threshold values of the financial integration measure. 7

For computational and econometric issues relating least square estimation and the non-standard asymptotic theory of inference refer to Hansen (1999).

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

6

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

5. Results I use the basic specification (7), with threshold variables given by measures of financial integration. Specifically, I use (1) equity volume: total FDI and FPI positions (relative to GDP), and (2) total volume: total positions (relative to GDP). A summary of results is presented in Table A.4. Each column gives results for two different specification, which are discussed in detail below. In all cases, the third threshold is not significant, and therefore Table A.4 only specifies the estimates corresponding to the first two computed thresholds.8 5.1. Thresholds based on total FDI and FPI To read the results, start by looking at the estimation results for the full sample (first column of Table A.4). Here, two significantly different threshold values of equity volume to GDP ratio are identified. Threshold 1 has a point estimate of ≈ 0.435 = 43.5.9% and a 95% confidence interval of ≈ [0.085, 0.435]. Threshold 2 has a point estimate of ≈ 0.841 = 84.1% and a 95% confidence interval of ≈ [0.841, 0.878]. The F-test for a single threshold and double thresholds are highly significant with bootstrap p-values of 0.00 and 0.02, respectively. The third threshold value, which is not reported in Table A.4, has a point estimate of ≈ 0.116 = 11.6% and a 95% confidence interval of ≈ (0.085, 0.274). However, the bootstrap p-value for the triple threshold regression is not significant at the 5% level, and is therefore indicated by X in Table A.4. The confidence interval for Threshold 1 is not very tight; however, it does not overlap with the 95% confidence interval of Threshold 2, and so the estimation indicates the existence of three robust regimes, with different coefficients of risk-sharing: Regime 1 lies below Threshold 1 and has an estimated risk-sharing coefficient of (1 − β1c ) ≈ 0.171 = 17.1 %, which (with a whiteerrorof ≈ 0.05) is statistically different from 1 but economically small, suggesting some limited risk-sharing. Regime 2 lies between Thresholds 1 and 2 and has an estimated risk-sharing coefficient of (1 − β2c ) ≈ −0.100. The estimated coefficient is significantly different from 1 but not significantly different from 0. This implies that there is no consumption risksharing in this regime. In addition, relative to Regime 1, Regime 2 indicates the potential presence of excess volatility (correlation between country-specific consumption growth and country-specific output growth).9 Regime 3 lies above Threshold 2 and has an estimated risk-sharing coefficient of (1 − β3c ) ≈ 0.50, which is significantly greater than 0 and smaller than 1. Although perfect risk-sharing is rejected, country–time pairs in Regime 3 do exhibit significant risksharing, with an average correlation of country-specific consumption to output growth of approximately 50%. Fig. B.1 provides a graphical representation of the findings. Using the estimates of equity volume to GDP as the threshold levels, the sample is split and the scatter plot along with ordinary least square fitted lines for each regime (Regime 1, Regime 2, and Regime 3) is plotted in the first three panels. Risk-sharing patterns from the estimates of β coefficients are also evident from the fourth panel of Fig. B.1 where the slope estimates are superimposed in a single illustration for the threshold variables. Fig. B.2 facilitates comparison of these estimated regime-dependent coefficients of consumption risk sharing with a reference line for no risk sharing and full risk sharing.10 Finally, Tables A.5, A.6, and A.7 report the number of countries in each regime at each date, and Tables A.11, A.12, and A.13 inAppendix A report the years in which each country (in the sample) falls into each of the regimes. About 134 country–year pairs (mostly emerging and developing countries) fall into Regime 2 for some part of their financial-integration process. In addition, only a small number of emerging economies have successfully transitioned into Regime 3 (e.g., Chile, Israel, Jordan, Malaysia, Singapore, and South Africa). Except for Singapore, the rest of the emerging economies only transitioned into Regime 3 after 1999. Furthermore, the idea of the destabilizing feature of Regime 2 is corroborated from the observation that many emerging and developing economies move from Regime 1 to Regime 2 (for example, Bolivia, Zimbabwe, Argentina, Brazil, Korea, Thailand, and Venezuela) and then move back to Regime 1. This transitory nature of Regime 2 captures the idea that the cost associated with financial integration in Regime 2 is a hurdle for many economies; therefore, the economies transition back into Regime 1 when they are unable to overcome the cost. For the very same reason of transitory nature, Regime 2 is narrow, i.e., contains only a few country– year pairs. Advanced countries fall into Regime 3 overwhelmingly for a substantial number of years in the sample period (especially in the recent years). In addition, for advanced economies only Greece and Japan fall back from Regime 2 to Regime 1, during the global financial crisis. 5.2. Thresholds based on total foreign assets and liabilities Analogous to column 1 of Table A.4, column 2 reports estimation results for the full sample using total volume of holdings to GDP as the threshold variable. For this sample, the Threshold 1 point estimate is ≈ 1.43 = 134% with a 95% confidence interval 8

I omit a complete breakdown of the results of four iterations to conserve space. Additional appendix containing complete tables is available upon request. Alternatively, β2c is not significantly different from 1, indicating that Regime 2 has no evidence of consumption risk sharing. 10 Confidence intervals at the α significance level for β coefficients can be found using (β ± tα ∗ (HETS. E)), where tα is the t statistics for α level of significance and HETS.E is the heteroskedasticity-adjusted white standard errors. 9

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

7

of ≈ [0.944, 1.73], and the Threshold 2 point estimate is ≈ 3.01 = 301% with a 95% confidence interval of ≈ [3.01, 3.01]. Because the two confidence intervals for estimates on the sample do not overlap, the F-test for the null hypothesis of equality of coefficients (no threshold) is rejected, and the corresponding p value is 0.00, we can again identify three robust regimes, with different coefficients of risk-sharing: Regime 1 lies below Threshold 1 and has an estimated risk-sharing coefficient of (1 − βc1) ≈ 0.21, which (with a whiteerrorof ≈ 0.052) is statistically and economically different from 0 and 1, suggesting almost 21% risk-sharing. Regime 2 lies between Threshold 1 and Threshold 2 and has an estimated risk-sharing coefficient of (1 − βc2) ≈ 0 = 0 %, which is statistically different from 1 but not 0 (alternatively, β2c is significantly different from 0 but not from 1), suggesting no significant risk-sharing (or perfect co-movement of domestic consumption growth rate and output growth rate). Regime 3 lies above Threshold 2 and has an estimated risk-sharing coefficient of (1 − βc3) ≈ 0.50, which is significantly less than 1 (but greater than 0). In this regime, country–time pairs achieve an average of approximately 50% consumption risk-sharing. Because the coefficients are similar, Fig. B.1 also illustrates the scatter plot and OLS-fitted line for the three sub-samples that use the split based on the estimates of threshold levels of total volume of capital holdings (normalized to GDP). Risk-sharing patterns from the estimates of β coefficients are presented in the fourth panel of Fig. B.1 and B.2, where the slope estimates are superimposed in a single illustration for the two threshold variables. The reference lines in Fig. B.2 facilitate comparison between β = 1, which captures no risksharing whereas β = 0 captures perfect risk-sharing. Regime 1 and Regime 3 estimated coefficients are clearly between 0 and 1, whereas Regime 2's estimated coefficient is ≈ 1. Tables A.8, A.9, and A.10 indicate the number of countries in each regime for each time period, and Tables A.14, A.15, and A.16 in Appendix A give a more detailed distribution of country–year pairs in the three identified regimes. Qualitatively, these results are similar to the results achieved in Section 5.1. A significant number of emerging and developing economies (about 222 country–year pairs) fall into Regime 2 especially in recent years, while only a small number of emerging and developing economies have transitioned into Regime 3 in the sample period (e.g., Jordan and Singapore). The destabilizing Regime 2 is more prominent using this measure of financial integration. More emerging and developing economies transition out of Regime 1 to move to Regime 2 but transition back to Regime 1 because of the overwhelming costs of integration (in terms of relative reduced consumption risk sharing). Again, advanced countries overwhelmingly fall into Regime 3 over the full sample period. One key difference between the quantitative results from these two measures is that the threshold estimated from total volume is stricter (i.e., much higher threshold) than the threshold estimated from equity volume because of the inclusion of debt stock in the total volume measure. Inclusion of debt stock (due to the pro-cyclicality issues) pushes the threshold, beyond which an economy graduates into Regime 3, to the right. Larger threshold is the primary reason why some economies which graduated into regime 3 (for some recent years) under equity volume financial integration measure do not graduate into Regime 3 under total volume financial integration measure. However, the number of such country–year sample is very small relative to the total sample and does not inhibit the robustness of the results provided in this section. The difference is in part due to the inclusion of more recent years (2007–2009), which captures the period of the global financial crisis. The global financial crisis presents a unique period in which the global output fell, coupled with increased volatility in the financial sector. Economies relying more on debt stock were exposed to debt-reversal or had to pay risk premia due to the potential possibility of default. Therefore, the threshold in the total volume financial integration measure is higher, and only a few emerging and developing country–year pairs fall into Regime 3.11 6. Summary and discussion 6.1. Summary To summarize the findings in Section 5, there is significant evidence of threshold effects of financial integration on cross-border consumption risk-sharing. Specifically, data suggest the presence of two thresholds and three corresponding regimes in which risksharing patterns differ significantly. Regime 1 (characterized by low levels of financial integration) has a statistically positive but economically small degree of consumption risk-sharing. Regime 3 (characterized by high levels of financial integration) has a statistically and economically significant consumption risk-sharing. However, at the intermediate levels of financial integration, consumption risk-sharing is insignificant, suggesting a more volatile intermediate phase relative to the other regimes in the process of financial integration. Confidence intervals around the threshold levels accompany the threshold estimates. These confidence regions are fairly tight, and results remain significant across the confidence regions. Qualitative results are also robust to alternative measures of financial integration. The estimation of tight confidence intervals is important for making policy conclusions because one implication of the results is that the effect of financial integration on consumption risk-sharing is non-linear, with the second regime showing no benefit in terms of consumption risk-sharing. Such a regime can potentially be destabilizing for fragile economies, indicating potential upfront costs of financial integration that are turned into benefits only after integration into international capital markets has proceeded far enough. Threshold estimate of Regime 2 accompany tight confidence intervals and indicate that passing the intermediate regime 2 requires a 11 The analysis (based on these two measures of financial integration — equity volume and total volume) that excludes this time period (2007–2009) from the sample period exhibits results that are quantitatively and qualitatively even closer.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

8

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

large increase in financial-integration measure for example, of 40% of gross FDI and FPI relative to GDP. Therefore, the intermediate regime may hinder economies' ability to achieve international consumption risk-sharing through financial integration. To provide a theoretical context for the findings, in Section 6.2, I investigate the underlying differences between the compositions of capital in the country–year pairs in each of the regimes. 6.2. Discussion The underlying difference between the degrees of risk-sharing in each regime can be understood by observing the average composition of capital stock (such as FDI, FPI, and debt as a percentage of total capital stock) in each of the documented regimes. Fig. B.3–B.5 illustrate two broad observations. First, Fig. B.3 illustrates that the average debt assets and liabilities (as a percentage of total capital stock) monotonically decline from Regime 1 to Regime 3 whereas Fig. B.4 illustrates that the average foreign portfolio assets and liabilities (as a percentage of total capital stock) monotonically increase from Regime 1 to Regime 3. Second, Fig. B.5 illustrates that the average foreign direct investment assets (as a percentage of total stock) increase monotonically but foreign direct investment liabilities follow a non-monotonic path from Regime 1 to Regime 3 (with the highest FDI liabilities in Regime 2). Combining these observations with the theoretical underpinnings reconcile the varying degree of risk-sharing estimated across the three regimes. Limited risk-sharing in Regime 1 is mainly driven by the high ratio of debt stock (as a percentage of total stock): Debt stock can be utilized to smooth consumption; however, empirical studies have shown (see for e.g., Kose & Prasad, 2010) that debt stock is pro-cyclical and therefore not an effective form of risk-sharing. On average, country–year pairs in Regime 1 are dominated by debt stock (along with insignificant FDI and FPI capital stocks) and this is the primary reason behind the limited risk-sharing observed in Regime 1. Lack of risk-sharing in Regime 2 is mainly driven by primitive increases in FDI liabilities (as a percentage of total stock): Relative to Regime 1, Regime 2 experiences a reduction in debt stocks but a substantial increase in FDI liabilities. Until recently, FDI liabilities were considered a more stable form of capital. One of the key features of FDI is that the investors are directly involved in the operation of the enterprise (see e.g., Albuquerque, 2003). This feature eliminates the issue of inalienability. Such arguments imply that foreign direct investments may provide a better form of consumption risk-sharing. However, theoretical predictions (see e.g., Acemoglu & Zilibotti, 1997; Razin & Sadka, 2001) and empirical findings (see e.g., Neumann, Penl, & Tanku, 2009) support a counter argument for the primitive FDI liabilities. In particular, initial FDI liabilities may induce additional volatility in the economy and therefore may lead to a reduction in consumption risk-sharing. Country–year pairs in Regime 2 experience an initial increase in FDI liabilities which can potentially be driving the observed lack of consumption risk-sharing. This lack of consumption risk-sharing also points toward the destabilizing nature of Regime 2 which therefore, encompasses an additional and a significant cost of financial integration process.12 A substantial degree of risk-sharing in Regime 3 is mainly driven by increasing FPI stocks (as a percentage of total stock): Relative to Regime 1 and Regime 2, Regime 3 has on average highest FPI stock (as a percentage of total capital stock). Moreover, relative to regime 2, on average FDI liabilities (as a percentage of total capital stock) also decrease. Increase in FPI and FDI assets act as counter projects to diversify the additional volatility due to primitive increases in FDI liabilities accumulated in Regime 2. 7. Conclusion This paper has outlined an empirical strategy for identifying threshold effects of financial integration on consumption risk-sharing, based on the threshold identification method in Hansen (1999). Applied to a panel data set of 64 developing, emerging, and advanced economies from 1985–2007, I identify a robust qualitative feature of the data: consumption risk-sharing is negligible at low levels of financial integration, and significant (though imperfect) at high levels of financial integration. However, intermediate to these estimated thresholds is a regime with significantly worse consumption risk-sharing (and some evidence of a regime with excess volatility). Conflicting evidence on the relationship between financial integration and consumption risk-sharing from the empirical literature can be understood better in the light of the findings reported in this paper. Specifically, ignoring the intermediate regime, or trying to impose a linear structure on the data-generating process is liable to result in misleading conclusions about the potential benefits of financial integration for consumption risk-sharing. In fact, the results suggest that financial integration is associated with significant consumption risk-sharing, but only after a potentially destabilizing transitory phase. Acknowledgments I am grateful to Nancy Chau, Eswar Prasad, and Viktor Tysernnikov for insightful discussions. I also thank David Cesarini, Chetan Dave, Jean Imbs, John Leahy, Maximilian Mihm, Romain Ranciere, and seminar participants at NYUAD for comments, and Bent Sorensen for providing data for replication purposes. 12 Financial integration is a process that encompasses short-run costs with long-run benefits (See for e.g., Agenor, 2003). In the current paper, I capture an additional short-run cost in a reduced form through the intermediate regime in the process of financial integration (Regime 2). The gap between Regime 1 to Regime 3 is substantial; therefore we observe some emerging and developing economies move briefly to Regime 2 and then either jump back or forward to Regime 1 or Regime 3, respectively.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

9

Appendix A. Tables

Table A.1 List of countries. Country type

Emerging

Advanced

Developing

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

Argentina Brazil Chile China Colombia Costa Rica Egypt India Indonesia Israel Jordan Korea Malaysia Mexico Pakistan Peru Philippines Singapore South Africa Thailand Turkey Venezuela

Australia Austria Belgium Canada Denmark Finland France Germany Greece Ireland Italy Japan Netherlands New Zealand Norway Portugal Spain Sweden Switzerland United Kingdom United States

Algeria Bolivia Cameroon Dominican R. Ecuador El Salvador Ghana Guatemala Honduras Jamaica Mauritius Nepal Nicaragua Paraguay Senegal Sri Lanka Trinidad and Tobago Tunisia Uruguay Zimbabwe

Note: The sample comprises 64 countries — 21 developing economies, 22 emerging economies, and 22 advanced economies.

Table A.2 Data sources. Variable

Source

Stock of external liabilities Stock of external assets Stock of FDI liabilities Stock of equity liabilities Stock of external debt liabilities Stock of FDI assets Stock of equity assets Stock of external debt assets GDP GDP per capita GNI per capita Consumption per capita

EWN II EWN II EWN II EWN II World Bank EWN II EWN II EWN II WDI-WB WDI-WB WDI-WB WDI-WB

Note: WDI-WB: World Development Indicators-World Bank, EWNII: Updated External Wealth of Nations. All data from EWN II and WDI are in constant (2005) price US dollars.

Table A.3 Summary statistics. Stats

GDP per capita

Consumption per capita

Total volume to GDP

Equity volume to GDP

p10 p25 p50 p75 p90

619.63 1297.92 3752.15 17,970.62 26,065.36

483.43 1042.90 2940.30 13,724.87 18,991.67

0.618 0.857 1.226 1.946 3.796

0.068 0.121 0.276 0.665 1.131

Note: This table provides summary statistics in terms of percentiles. px corresponds to the xth percentile. 50th percentile (p50) corresponds to the median statistics.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

10

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx Table A.4 Threshold regression results.

Threshold 1 CI min CI max Threshold 2 CI min CI max Threshold 3 CI min CI max F Bootstrap P βo OLS SE HET SE β1c OLS SE HET SE β2c OLS SE HET SE β3c OLS SE HET SE β4c OLS SE HET SE Total observations: Number of years number of countries

FDIþ FPI GDP

Total GDP

0.435 0.085 0.435 0.841 0.841 0.878 X X X 11.1 0.020 0.00360⁎⁎⁎ 0.00429 0.00792 0.829⁎⁎⁎ 0.0330 0.0555 1.10 0.0767 0.150 0.503⁎⁎⁎ 0.101 0.159 X X X 1536 25 64

1.43 0.944 1.729 3.015 3.015 3.015 X X X 12.3 0.006 0.00222⁎⁎⁎ 0.00423 0.00806 0.793⁎⁎⁎ 0.0377 0.0520 1.01 0.050 0.107 0.506⁎⁎⁎ 0.0948 0.256 X X X 1536 25 64

Note: The results (estimates of threshold levels and point estimates of (1) time trend (βo) (2) coefficients of comovement between country-specific consumption and output growth rates (β1c , β2c , β3c , β4c )) provided in this table are based on panel regressions using specification 7 with (1) equity volume (normalized to GDP) and (2) total volume of capital holdings (normalized to GDP) as the threshold variable in column 1 and column 2, respectively. Corresponding to threshold level, confidence intervals (CI min and CI max) are also provided. OLS (OLS SE) and white, heteroskedastic corrected standard errors (HET SE) are provided beneath each estimate of the coefficients. Results significantly different from 1 are reported using the following symbols: p b 0.01, * *p b 0.05, and *p b 0.1. For details of regression specification, see Section 4.3.

Table A.5 Distribution of developing economies: FDI & FPI threshold. Developing

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Total

Regime 1 Regime 2 Regime 3 Total

20 1 0 21

19 2 0 21

19 2 0 21

19 2 0 21

19 2 0 21

18 3 0 21

18 3 0 21

19 2 0 21

19 1 1 21

19 1 1 21

19 1 1 21

19 1 1 21

19 1 1 21

19 1 1 21

18 2 1 21

18 2 1 21

18 2 1 21

16 3 2 21

16 3 2 21

16 3 2 21

15 4 2 21

15 5 1 21

14 6 1 21

16 4 1 21

15 4 2 21

442 61 22 525

Note: This table shows the distribution of developing countries in each year falling in each regime. Regimes are identified and estimated using specification 7 with equity volume of capital holdings (normalized to GDP) as the threshold variable. The sample contains 21 developing countries. The last column provides the total number of observations in the sample period (1985–2007), falling in each regime. For a more detailed country-specific distribution, see Table A.11.

Table A.6 Distribution of emerging economies: FDI & FPI threshold. Emerging

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Total

Regime 1 Regime 2 Regime 3 Total

21 0 1 22

21 0 1 22

21 0 1 22

21 0 1 22

21 0 1 22

21 0 1 22

21 0 1 22

20 1 1 22

20 1 1 22

19 2 1 22

19 2 1 22

19 2 1 22

18 3 1 22

18 3 1 22

15 3 4 22

16 3 3 22

16 4 2 22

14 5 3 22

12 7 3 22

12 6 4 22

11 7 4 22

10 6 6 22

10 6 6 22

13 5 4 22

9 7 6 22

418 73 59 550

Note: This table shows the distribution of emerging countries in each year falling in each regime. Regimes are identified and estimated using specification 7 with equity volume of capital holdings (normalized to GDP) as the threshold variable. The sample contains 22 emerging countries. The last column provides the total number of observations in the sample period (1985–2007), falling in each regime. For a more detailed country-specific distribution, see Table A.12.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

11

Table A.7 Distribution of advanced economies: FDI & FPI threshold. Advanced

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Total

Regime 1 Regime 2 Regime 3 Total

16 4 1 21

15 5 1 21

14 6 1 21

13 7 1 21

13 6 2 21

13 6 2 21

12 7 2 21

12 7 2 21

11 4 6 21

11 4 6 21

11 3 7 21

9 5 7 21

7 5 9 21

3 8 10 21

3 6 12 21

2 5 14 21

2 5 14 21

2 4 15 21

2 3 16 21

2 2 17 21

1 2 18 21

0 2 19 21

0 2 19 21

2 3 16 21

1 2 18 21

177 113 235 525

Note: This table shows the distribution of advanced countries in each year falling in each regime. Regimes are identified and estimated using specification 7 with equity volume of capital holdings (normalized to GDP) as the threshold variable. The sample contains 22 advanced countries. The last column provides the total number of observations in the sample period (1985–2007), falling in each regime. For a more detailed country specific distribution, see Table A.13.

Table A.8 Distribution of developing economies: Total threshold. Developing

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Total

1 2 3 Total

19 2 0 21

18 3 0 21

16 5 0 21

18 2 1 21

15 5 1 21

15 5 1 21

17 3 1 21

17 3 1 21

17 3 1 21

16 4 1 21

16 4 1 21

17 4 0 21

17 4 0 21

17 4 0 21

15 6 0 21

14 7 0 21

15 5 1 21

12 7 2 21

13 6 2 21

12 8 1 21

14 6 1 21

14 6 1 21

14 6 1 21

15 5 1 21

13 7 1 21

386 120 19 525

Note: This table shows the distribution of developing countries in each year falling in each regime. Regimes are identified and estimated using specification 7 with total volume of capital holdings (normalized to GDP) as the threshold variable. The sample contains 21 developing countries. The last column provides the total number of observations in the sample period (1985–2007), falling in each regime. For a more detailed country-specific distribution see the Table A.15.

Table A.9 Distribution of emerging economies: Total threshold. Emerging

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Total

Regime 1 Regime 2 Regime 3 Total

18 3 1 22

17 4 1 22

17 4 1 22

17 4 1 22

18 3 1 22

18 3 1 22

18 2 2 22

18 3 1 22

17 4 1 22

17 4 1 22

18 3 1 22

18 3 1 22

19 2 1 22

17 4 1 22

14 7 1 22

16 4 2 22

17 3 2 22

15 5 2 22

15 5 2 22

15 5 2 22

16 4 2 22

14 6 2 22

14 6 2 22

17 4 1 22

14 7 1 22

414 102 34 550

Note: This table shows the distribution of emerging countries in each year falling in each regime. Regimes are identified and estimated using specification 7 with total volume of capital holdings (normalized to GDP) as the threshold variable. The sample contains 22 emerging countries. The last column provides the total number of observations in the sample period (1985–2007), falling in each regime. For a more detailed country-specific distribution, see Table A.15.

Table A.10 Distribution of advanced economies: Total threshold. Advanced

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Total

1 2 3 Total

14 4 3 21

16 2 3 21

15 3 3 21

15 3 3 21

14 4 3 21

15 3 3 21

13 5 3 21

14 4 3 21

10 7 4 21

10 7 4 21

10 7 4 21

9 7 5 21

5 11 5 21

3 13 5 21

3 11 7 21

3 9 9 21

3 7 11 21

1 8 12 21

1 7 13 21

1 7 13 21

0 8 13 21

0 7 14 21

0 7 14 21

0 6 15 21

0 5 16 21

175 162 188 525

Note: This table shows the distribution of advanced countries in each year falling in each regime. Regimes are identified and estimated using specification 7 with total volume of capital holdings (normalized to GDP) as the threshold variable. The sample contains 22 advanced countries. The last column provides the total number of observations in the sample period (1985–2007), falling in each regime. For a more detailed country-specific distribution, see Table A.16.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

12

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

Table A.11 Distribution of developing economies for FDI & FPI holdings. Advanced

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Algeria Bolivia Cameroon Costa Rica Dominican Republic Ecuador El Salvador Ghana Guatemala Honduras Mauritius Nicaragua Papua New Guinea Paraguay Senegal Sri Lanka Syria Trinidad and Tobago Tunisia Uruguay Zimbabwe

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 2 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 2 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2 1 1

1 2 1 1 1 1 1 1 1 1 3 2 1 1 1 1 1 3 2 1 1

1 2 1 1 1 1 1 1 1 1 3 2 1 1 1 1 1 3 2 1 1

1 2 1 1 1 1 1 1 1 1 3 2 1 1 1 1 1 3 2 1 1

1 2 1 1 1 1 1 1 1 2 3 2 1 1 1 1 1 3 2 1 1

1 2 1 1 1 1 1 1 1 2 3 2 1 1 1 1 1 2 2 1 1

1 2 1 1 1 1 1 1 1 2 3 2 1 1 1 1 1 2 2 1 2

1 1 1 1 1 1 1 1 1 2 3 2 1 1 1 1 1 2 2 1 1

1 1 1 2 1 1 1 1 1 2 3 2 1 1 1 1 1 3 2 1 1

Note: This table provides the distribution of country–time pairs (for a sample of 21 developing countries) in each of the three identified regimes. Identification and estimation of regimes is based on equity volume as the threshold variable. For details of identification and estimation of threshold levels and corresponding regimes, see Sections 4 and 5.

Table A.12 Distribution of emerging economies for FDI & FPI holdings. Emerging

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Argentina Brazil Chile China Colombia Egypt India Indonesia Israel Jordan Korea Malaysia Mexico Morocco Pakistan Peru Philippines Singapore South Africa Thailand Turkey Venezuela

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 3 1 1 1 1

1 1 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 3 1 1 1 1

1 1 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 3 1 1 1 1

1 1 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 3 1 1 1 1

1 1 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 3 2 1 1 1

1 1 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 3 2 1 1 1

1 1 3 1 1 1 1 1 2 1 1 3 1 1 1 1 1 3 3 2 1 2

1 1 3 1 1 1 1 1 2 2 1 2 1 1 1 1 1 3 3 1 1 1

1 1 3 1 1 1 1 1 2 2 1 2 1 1 1 1 1 3 2 1 1 1

2 1 3 1 1 1 1 1 1 2 1 2 1 1 1 1 1 3 3 2 1 2

2 2 3 1 1 1 1 1 2 2 1 2 1 1 1 1 1 3 3 2 1 2

2 2 3 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 3 2 1 2

2 2 3 1 1 1 1 1 2 3 2 2 2 1 1 1 1 3 3 2 1 1

2 2 3 1 1 1 1 1 3 3 2 3 2 2 1 1 1 3 3 2 1 1

1 2 3 1 1 1 1 1 3 3 2 3 2 2 1 2 1 3 3 2 1 1

1 1 3 1 1 1 1 1 2 3 1 2 1 2 1 2 1 3 3 2 1 1

1 2 3 1 2 1 1 1 3 3 2 3 2 2 1 2 1 3 3 2 1 1

Note: This table provides the distribution of country–time pairs (for a sample of 22 emerging countries) in each of the three identified regimes. Identification and estimation of regimes is based on equity volume as the threshold variable. For details of identification and estimation of threshold levels and corresponding regimes, see Sections 4 and 5.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

13

Table A.13 Distribution of advanced economies for FDI & FPI holdings. Advanced

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Australia Austria Belgium Canada Denmark Finland France Germany Greece Ireland Italy Japan Netherlands New Zealand Norway Portugal Spain Sweden Switzerland United Kingdom United States

1 1 1 1 1 1 2 1 1 2 1 1 2 1 1 1 1 1 3 2 1

1 1 2 1 1 1 2 1 1 2 1 1 2 1 1 1 1 1 3 2 1

2 1 2 2 1 1 1 1 1 2 1 1 2 1 1 1 1 1 3 2 1

2 1 2 2 1 1 2 1 1 2 1 1 2 1 1 1 1 1 3 2 1

2 1 2 2 1 1 2 1 1 2 1 1 3 1 1 1 1 1 3 2 1

2 1 2 2 1 1 1 1 1 2 1 1 3 2 1 1 1 1 3 2 1

2 1 2 2 1 1 2 1 1 2 1 1 3 2 1 1 1 1 3 2 1

2 1 2 2 1 1 2 1 1 2 1 1 3 2 1 1 1 1 3 2 1

2 1 3 2 1 1 2 1 1 3 1 1 3 3 1 1 1 2 3 3 1

2 1 3 2 1 1 2 1 1 3 1 1 3 3 1 1 1 2 3 3 1

2 1 3 3 1 1 2 1 1 3 1 1 3 3 1 1 1 2 3 3 1

2 1 3 3 1 1 2 1 1 3 1 1 3 3 2 1 1 2 3 3 2

2 1 3 3 2 2 3 1 1 3 1 1 3 3 2 1 1 3 3 3 2

2 1 3 3 2 3 3 2 1 3 2 1 3 3 2 2 2 3 3 3 2

3 1 3 3 3 3 3 2 1 3 2 1 3 3 2 2 2 3 3 3 2

3 2 3 3 3 3 3 3 1 3 2 1 3 3 2 2 3 3 3 3 2

3 2 3 3 3 3 3 3 1 3 2 1 3 2 3 2 3 3 3 3 2

3 2 3 3 3 3 3 3 1 3 2 1 3 3 3 2 3 3 3 3 2

3 2 3 3 3 3 3 3 1 3 2 1 3 3 3 3 3 3 3 3 2

3 3 3 3 3 3 3 3 1 3 2 1 3 3 3 3 3 3 3 3 2

3 3 3 3 3 3 3 3 1 3 2 2 3 3 3 3 3 3 3 3 3

3 3 3 3 3 3 3 3 2 3 3 2 3 3 3 3 3 3 3 3 3

3 3 3 3 3 3 3 3 2 3 3 2 3 3 3 3 3 3 3 3 3

2 3 3 3 3 3 3 3 1 3 2 1 3 2 3 3 3 3 3 3 3

3 3 3 3 3 3 3 3 1 3 2 2 3 3 3 3 3 3 3 3 3

Note: This table provides the distribution of country–time pairs (for a sample of 22 advanced countries) in each of the three identified regimes. Identification and estimation of regimes is based on equity volume as the threshold variable. For details of identification and estimation of threshold levels and corresponding regimes, see Sections 4 and 5.

Table A.14 Distribution of developing economies for total holdings. Developing

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Algeria Bolivia Cameroon Costa Rica Dominican Republic Ecuador El Salvador Ghana Guatemala Honduras Mauritius Nicaragua Papua New Guinea Paraguay Senegal Sri Lanka Syria Trinidad and Tobago Tunisia Uruguay Zimbabwe

1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 2 1

1 2 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 2 1

1 2 1 1 1 2 1 1 1 1 1 2 1 1 1 1 2 1 2 1 1

1 1 1 1 1 2 1 1 1 1 1 3 1 1 1 1 2 1 1 1 1

1 1 1 1 1 2 1 1 1 1 1 3 1 1 1 1 2 2 2 2 1

1 1 1 1 1 2 1 1 1 2 1 3 2 1 1 1 2 1 1 2 1

1 1 1 1 1 2 1 1 1 2 1 3 1 1 1 1 2 1 1 1 1

1 1 1 1 1 2 1 1 1 2 1 3 1 1 1 1 2 1 1 1 1

1 1 1 1 1 1 1 1 1 2 1 3 1 1 1 1 2 2 1 1 1

1 1 1 1 1 1 1 1 1 2 1 3 1 1 1 1 2 2 2 1 1

1 1 1 1 1 1 1 1 1 2 1 3 1 1 1 1 2 2 2 1 1

1 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 2 2 1 1 1

1 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 2 2 1 1 1

1 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 2 2 1 1 1

1 2 1 1 1 2 1 1 1 2 1 2 1 1 1 1 2 2 1 1 1

1 2 1 1 1 2 1 2 1 1 1 2 1 2 1 1 2 2 1 1 1

1 2 1 1 1 1 1 2 1 1 1 2 1 3 1 1 1 2 1 2 1

1 2 1 1 1 1 1 2 1 1 3 2 1 3 2 1 1 2 2 2 1

1 2 1 1 1 1 1 2 1 1 3 2 1 3 1 1 1 2 2 2 1

1 2 1 1 1 1 1 2 1 2 3 2 1 2 1 1 1 2 2 2 1

1 2 1 1 1 1 1 1 1 1 3 2 1 2 1 1 1 2 2 2 1

1 1 1 1 1 1 1 1 1 1 3 2 1 2 1 1 1 2 2 2 2

1 1 1 1 1 1 1 1 1 1 3 2 1 2 1 1 1 2 2 2 2

1 1 1 1 1 1 1 1 1 1 3 2 1 1 1 1 1 2 2 2 2

1 1 1 1 1 1 1 1 1 1 3 2 2 2 1 1 1 2 2 2 2

Note: This table provides the distribution of country–time pairs (for a sample of 21 developing countries) in each of the three identified regimes. Identification and estimation of regimes is based on total volume of capital holdings as the threshold variable. For details of identification and estimation of threshold levels and corresponding regimes, see Sections 4 and 5.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

14

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

Table A.15 Distribution of emerging economies for total holdings. Emerging

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Argentina Brazil Chile China Colombia Egypt India Indonesia Israel Jordan Korea Malaysia Mexico Morocco Pakistan Peru Philippines Singapore South Africa Thailand Turkey Venezuela

1 1 2 1 1 2 1 1 2 1 1 1 1 1 1 1 1 3 1 1 1 1

1 1 2 1 1 2 1 1 2 1 1 2 1 1 1 1 1 3 1 1 1 1

1 1 2 1 1 2 1 1 2 1 1 2 1 1 1 1 1 3 1 1 1 1

1 1 2 1 1 2 1 1 1 2 1 1 1 1 1 2 1 3 1 1 1 1

1 1 1 1 1 2 1 1 1 2 1 1 1 1 1 1 1 3 1 1 1 2

1 1 1 1 1 2 1 1 1 2 1 1 1 1 1 1 1 3 1 1 1 2

1 1 1 1 1 2 1 1 1 3 1 1 1 1 1 1 1 3 1 1 1 2

1 1 1 1 1 2 1 1 1 2 1 1 1 1 1 1 1 3 1 1 1 2

1 1 1 1 1 2 1 1 1 2 1 2 1 1 1 1 1 3 1 1 1 2

1 1 1 1 1 2 1 1 1 2 1 2 1 1 1 1 1 3 1 1 1 2

1 1 1 1 1 2 1 1 1 2 1 2 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 1 3 1 1 1 2

1 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 1 3 1 1 1 1

1 1 1 1 1 1 1 2 1 2 1 2 1 1 1 1 1 3 1 2 1 1

1 1 2 1 1 1 1 2 2 2 1 2 1 1 1 1 1 3 2 2 1 1

1 1 2 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 2 1 1 1

1 1 2 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 1 1 1 1

2 1 2 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 1 1 1 2

2 1 2 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 1 1 1 2

2 1 2 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 1 1 1 2

2 1 2 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 1 1 1 1

2 1 2 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 2 2 1 1

2 1 2 1 1 1 1 1 2 3 1 2 1 1 1 1 1 3 2 2 1 1

1 1 2 1 1 1 1 1 2 2 1 2 1 1 1 1 1 3 1 1 1 1

1 1 2 1 1 1 1 1 2 2 2 2 1 1 1 1 1 3 2 2 1 1

Note: This table provides the distribution of country–time pairs (for a sample of 22 emerging countries) in each of the three identified regimes. Identification and estimation of regimes is based on total volume of capital holdings as the threshold variable. For details of identification and estimation of threshold levels and corresponding regimes, see Sections 4 and 5.

Table A.16 Distribution of advanced economies for total holdings. Advanced

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07

08

09

Australia Austria Belgium Canada Denmark Finland France Germany Greece Ireland Italy Japan Netherlands New Zealand Norway Portugal Spain Sweden Switzerland United Kingdom United States

1 2 3 1 2 1 1 1 1 2 1 1 2 1 1 1 1 1 3 3 1

1 1 3 1 1 1 1 1 1 2 1 1 2 1 1 1 1 1 3 3 1

1 1 3 1 2 1 1 1 1 2 1 1 2 1 1 1 1 1 3 3 1

1 1 3 1 2 1 1 1 1 2 1 1 2 1 1 1 1 1 3 3 1

1 1 3 1 2 1 2 1 1 2 1 1 2 1 1 1 1 1 3 3 1

1 1 3 1 2 1 1 1 1 2 1 1 2 1 1 1 1 1 3 3 1

1 1 3 1 2 1 2 1 1 2 1 1 2 1 1 1 1 2 3 3 1

1 1 3 1 2 1 1 1 1 2 1 1 2 2 1 1 1 1 3 3 1

1 1 3 2 2 2 2 1 1 3 1 1 2 2 1 1 1 2 3 3 1

1 1 3 2 2 2 2 1 1 3 1 1 2 2 1 1 1 2 3 3 1

1 2 3 2 2 1 2 1 1 3 1 1 2 2 1 1 1 2 3 3 1

1 2 3 2 2 1 2 1 1 3 1 1 3 2 1 2 1 2 3 3 1

1 2 3 2 2 2 2 2 1 3 2 1 3 2 2 2 1 2 3 3 1

2 2 3 2 2 2 2 2 1 3 2 1 3 2 2 2 2 2 3 3 1

2 2 3 2 2 3 2 2 1 3 2 1 3 2 2 2 2 3 3 3 1

2 2 3 2 3 3 3 2 1 3 2 1 3 2 2 2 2 3 3 3 1

2 3 3 2 3 3 3 2 1 3 2 1 3 2 2 3 2 3 3 3 1

2 3 3 2 3 3 3 3 2 3 2 1 3 2 2 3 2 3 3 3 2

2 3 3 2 3 3 3 3 2 3 2 1 3 2 3 3 2 3 3 3 2

2 3 3 2 3 3 3 3 2 3 2 1 3 2 3 3 2 3 3 3 2

2 3 3 2 3 3 3 3 2 3 2 2 3 2 3 3 2 3 3 3 2

2 3 3 2 3 3 3 3 2 3 2 2 3 2 3 3 3 3 3 3 2

2 3 3 2 3 3 3 3 2 3 2 2 3 2 3 3 3 3 3 3 2

2 3 3 2 3 3 3 3 2 3 2 2 3 2 3 3 3 3 3 3 3

3 3 3 2 3 3 3 3 3 3 2 2 3 2 3 3 3 3 3 3 2

Note: This table provides the distribution of country–time pairs (for a sample of 22 advanced countries) in each of the three identified regimes. Identification and estimation of regimes is based on total volume of capital holdings as the threshold variable. For details of identification and estimation of threshold levels and corresponding regimes, see Sections 4 and 5.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

15

Appendix B. Illustrations

-.4

-.2

0

.2

.4

.4 .2 0 -.2 -.2

-.1

0

.1

.2

-.2

-.1

0

.1

.2

.4

.4 .2

Regime 1, Regime 2, Regime 3

.2

Regime 3

Regime 2

-.2

0

Regime 3

-.4

-.2

0

Regime 1

-.4

Fitted Values: Country SpecificConsumption growth

Regime 2

-.4

Fitted Values: Country Specific Consumption growth

Regime 1

-.2

-.1

0 Country Specific GDP growth

.1

.2

-.2

-.1

0

.1

.2

Country Specific GDP growth

Fig. B.1. FDI & FPI holdings (or) total holdings: Regimes. Note: Panel (1,1), Panel (1,2), and Panel (2,1) plot ordinary least square fitted lines for Regime 1, 2, and 3 respectively. These scatter plots are superimposed in Panel (2,2). Identification and estimation of regimes are based on equity volume (normalized to GDP) as a threshold variable.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

Reference line: Beta = 1

Regime 2: Beta =1

.1

Regime 1: Beta < 1

0

Regime 3: Beta < 1

-.1

Reference line: Beta = 0

-.2

Fitted values: Country Specific Consumption growth

.2

16

-.2

-.1

0

.1

.2

Country specific GDP growth Regime 1

Regime 2

Regime 3

Fig. B.2. FDI & FPI holdings (or) total holdings: Regimes. Note: This figure reproduces Panel (2, 2) of Fig. B.1 with additional reference lines. β = 0 (β = 1) is associated with full risk-sharing (no risk-sharing) and its corresponding reference line is the horizontal axis, y = 0 (45° line, y = x). The gray line indicates Regime 1, the blue line indicates Regime 2, and the red line indicates Regime 3's OLS fitted lines. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

.15

.2

.25

.3

.35

17

.1

DEBT ASSETS (% of Total Capital Stock)

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

1985

1990

1995

2000

2005

2010

year REGIME 1

REGIME 2

REGIME 3

.6 .5 .4 .3 .2

DEBT LIABILITIES (% of Total Capital Holdings)

(a) Regime1, 2, & 3: DEBT ASSETS (% of total capital stock)

1985

1990

1995

2000

2005

2010

year REGIME 1

REGIME 2

REGIME 3

(b) Regime 1, 2, & 3: DEBT LIABILITIES (% of total capital stock) Fig. B.3. Average composition of debt assets and liabilities. (a) Regimes 1, 2, & 3: DEBT ASSETS (% of total capital stock). (b) Regimes 1, 2, & 3: DEBT LIABILITIES (% of total capital stock). Notes: Fig. B.3(a) and (b) illustrate average debt assets and liabilities as a percentage of total capital stock in Regimes 1, 2, and 3.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

0

FPI ASSETS (% of Total Capital Holdings) .05 .1 .15

18

1985

1990

1995

2000

2005

2010

year REGIME 1

REGIME 2

REGIME 3

0

FPI LIABILITIES (% of Total Capital Stock) .05 .1 .15

(a) Regime 1, 2, & 3: FPI ASSETS (% of total capital stock)

1985

1990

1995

2000

2005

2010

year REGIME 1

REGIME 2

REGIME 3

(b) Regime 1 , 2, & 3: FPI LIABILITIES (% of total capital stock) Notes: Figure 4(a) and Figure 4(b) illustrate average FPI assets and liabilities as a percentage of total capital stock in Regime 1, 2, and 3.

Fig. B.4. Average composition of FPI assets and liabilities. (a) Regimes 1, 2, & 3: FPI ASSETS (% of total capital stock). (b) Regimes 1, 2, & 3: FPI LIABILITIES (% of total capital stock). Notes: Fig. B.4(a) and (b) illustrate average FPI assets and liabilities as a percentage of total capital stock in Regimes 1, 2, and 3.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

19

.02

FDI ASSETS (% of Total Capital Stock) .06 .1 .04 .08 .12

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

1985

1990

1995

2000

2005

2010

year REGIME 1

REGIME 2

REGIME 3

FDI LIABILITIES (% of Total Capital Stock) .2 .25 .1 .15 .3

(a) Regime 1, 2, & 3: FDI ASSETS (% of total capital stock)

1985

1990

1995

2000

2005

2010

year REGIME 1

REGIME 2

REGIME 3

(b) Regime 1, 2, & 3: FDI LIABILITIES (% of total capital stock) Notes: Figure 5(a) and Figure 5(b) illustrate average FDI assets and liabilities as a percentage of total capital stock in Regime 1, 2, and 3.

Fig. B.5. Average composition of FDI assets and liabilities. (a) Regimes 1, 2, & 3: FDI ASSETS (% of total capital stock). (b) Regimes 1, 2, & 3: FDI LIABILITIES (% of total capital stock). Notes: Fig. B.5(a) and (b) illustrate average FDI assets and liabilities as a percentage of total capital stock in Regimes 1, 2, and 3.

Appendix C. Comparison of results I also compare the results of the previous section to two alternative specifications that the existing literature has traditionally used (based on the data set employed in Section 5): (1) specification 4 based on a linear interaction of financial-integration, and (2) specification 7 that imposes the median levels of the financial integration measures to exogenously bifurcate the data into two sub-samples (rather than estimating threshold levels directly from the data). Appendix C.1. Linear interactions Table Appendix C.1 reports the results based on the linear interaction specification (specification 4) for the full sample (64 economies from 1985–2009) and also separately for different country groups (developing, emerging, and advanced). Again, the current Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

20

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

Table C.17 Threshold regression results: FDI & FPI and total holdings (normalized to GDP). C it

Full: Total

Full: FDI & FPI

DEV: Total

DEV: FDI & FPI

EME: Total

EME: FDI & FPI

ADV: Total

ADV: FDI & FPI

κo

0.828*** (0.0512) 0.00167 (0.00756) 0.00140

0.829*** (0.0513) 0.00206 (0.00746)

0.773*** (0.0967) 0.0130 (0.0131) −0.000500

0.774*** (0.0967) 0.0133 (0.0132)

0.904*** (0.0557) −0.00794 (0.00833) 0.0173

0.906*** (0.0572) −0.00860 (0.00918)

0.554*** (0.0463) −0.0207** (0.00957) 0.0209

0.556*** (0.0469) −0.0229** (0.00988)

κ1 Total

κ 2GDP

(0.00971) FDIþ FPI

κ 2 GDP

Observations R-squared

1600 0.346

(0.00959)

(0.0315)

(0.0234)

−0.00558

−0.00758

0.0477

(0.00795) 1600 0.346

(0.00655) 525 0.229

(0.0984) 550 0.533

525 0.228

550 0.534

0.0820 525 0.310

(0.0692) 525 0.313

Note: Regression results are based on specification 4. Columns 1, 3, 5, and 7 (2, 4, 6, and 8) use linear interaction term between total volume (equity volume) of capital holdings (normalized to GDP) and idiosyncratic output growth rates. Columns 1 and 2 provides results for the full sample, and columns 3 and 4, 5 and 6, and 6 and 7 provide results for developing, emerging, and advanced countries, respectively. Robust standard errors are shown in parentheses. ***p b 0.01, **p b 0.05, and *p b 0.1.

analysis uses two alternative de facto measures of financial integration (equity volume, and total volume of capital holdings, both normalized to GDP). The first and second column show the results for the full sample for each of the de facto measures for a fixed-effects panel regression. The remaining columns report results for each of the country groups in the following order: developing, emerging, and advanced. A negative coefficient κ2 indicates a greater degree of financial inte gration that is associated with higher levels of consumption risk-sharing. However, for the full sample, as well as the separate sample of country groups, none of the coefficients for the interaction FDIþ FPI Total term (κ 2GDP and κ 2 GDP ) is statistically significant (see Table C.17). In particular, financial integration appears to have no significant effect on consumption risk-sharing. The division into developing, emerging, and advanced economies is based on IMF classifications, which generally use income per capita as a criterion for division; therefore, it reflects an implicit prior assumption that consumption risk-sharing is related to per-capita income. However, we find little theoretical justification for such an assumption, in contrast to the strong theoretical prediction that financial integration should promote risk-sharing. The threshold specification 7 therefore estimates the effect of financial integration on consumption risk-sharing more directly, by splitting the sample based on actual measures of financial integration, and then showing that threshold effects exist. The results in Section 5 also provide a more nuanced picture. Regime 2, where consumption risk-sharing is lowest, not only contains emerging economies, but also some developing economies and advanced economies. Moreover, the regime has a transitory nature, with countries moving both in and out of this regime. Appendix C.2. Median based thresholds The previous literature has sometimes captured potential threshold effects by a simple median-value bifurcation of the sample: Δlogðcit Þ−Δlogðct Þ ¼ μ i þ βo t ðΔlog ðyit Þ−Δlog ðyt ÞÞ 1 þ βc ðΔlog ðyit Þ−Δlog ðyt ÞÞð F it ≤γ med Þ 2 þ βc ðΔlog ðyit Þ−Δlog ðyt ÞÞð F it Nγmed Þ;

C:1

where γmed is the median level of the financial-integration measure in the data, and 1 − β1c (respectively, 1 − β2c ) is a measure of the average consumption risk-sharing for observations that lie below (respectively, above) the median level of financial integration. I use the median value of the equity volume, as well as the total volume of holdings (as a ratio of GDP), to split the sample, and present corresponding results in column 1 and column 2 of Table C.18. The results indicate that significant but economically small consumption risk-sharing is achieved. However, the results show insignificant improvement in risk-sharing from financial integration, which is evident when we compare β1c = 0.883 and β2c = 0.768. The F test reveals that the two estimates are statistically, not significantly different from each other.13 Similarly, the estimates of the coefficient of GDP growth with total volume to GDP below and above the median (compare β1c = 0.809⁎ ⁎ ⁎ and β2c = 0.845⁎ ⁎ ⁎) also show significant but economically small consumption risk-sharing, however, the difference between these coefficients is not statistically significant, and I find no evidence of median based threshold regimes. An F-test cannot reject the hypothesis of the absence of a threshold effect.14

13 14

The F test corresponding to βc1 = βc2 cannot be rejected based on the following test statistics: F(1, 1597) = 1.38 and Prob N F = 0.2410. The F test corresponding to βc1 = βc2 cannot be rejected based on the following test statistics: F(1, 1597) = 0.12 and Prob N F = 0.7279.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004

S. Malik / International Review of Economics and Finance xxx (2015) xxx–xxx

21

Table C.18 Threshold regression results: Total holdings (normalized to GDP). Cit

FDI & FPI

Total

βo

0.00558 (0.00703) 0.883*** (0.0624) 0.768*** (0.0791)

0.00139 (0.00765)

β1c β2c β1c β2c Observations R-squared

1600 0.347

0.809*** (0.0556) 0.845*** (0.0836) 1600 0.346

Note: Regression results in columns 1 and 2 are based, respectively, on specification Appendix C.2 for equity volume and total volume (both normalized to GDP) as the threshold variable. Robust standard errors are shown in parentheses. ***p b 0.01, **p b 0.05, and *p b 0.1.

The results reported in Table C.18 indicate (1) a non-coherent picture from the two alternative measures of financial integration and (2) the absence of threshold effects. Comparing the results in Table C.18 with the results of Section 5 therefore indicates the danger of making policy conclusions through pre-imposed threshold levels. A close comparison of the results indicate the first threshold estimate of = 0.43 in Section 5 lies above γmed = 0.27, so that Regime 2 is entirely absorbed by the subsample with lower and higher financial integration in the median threshold model. Country–time pairs in Regime 2 therefore “pull down” the estimated consumption risk-sharing for countries falling above and below median financial integration, masking the non-linear effect of financial integration on consumption risk-sharing. Therefore, the estimated coefficients of the median threshold model are misleading as inputs into policy evaluation of the effects of financial integration on consumption risk-sharing. References Acemoglu, D., & Zilibotti, F. (1997). Was prometheus unbound by chance? Risk, diversification, and growth. Journal of Political Economy, 105, 709–751. Agenor, P.R. (2003). Benefits and costs of international financial integration: Theory and facts. The World Economy, 26, 1089–1118. Albuquerque, R. (2003). The composition of international capital flows: Risk sharing through foreign direct investment. Journal of International Economics, 61, 353–383. Ambler, S., Cardia, E., & Zimmermann, C. (2004). International business cycles: What are the facts? Journal of Monetary Economics, 51, 257–276. Backus, D.K., Kehoe, P.J., & Kydland, F.E. (1992). International real business cycles. Journal of Political Economy, 100, 745–775. Bai, Y., & Zhang, J. (2012). Financial integration and international risk sharing. Journal of International Economics, 86, 17–32. Canova, F., & Ravn, M.O. (1996). International consumption risk sharing. International Economic Review, 37, 573–601. Chinn, M., & Ito, H. (2008). A new measure of financial openness. Journal of Comparative Policy Analysis, 10, 309–322. Demyanyk, Y., Ostergaard, C., & Sorensen, B.E. (2008). Risk sharing and portfolio allocation in EMU. European Economy — Economic Papers 334Directorate General Economic and Monetary Affairs, European Commission. Edison, H.J., Klein, M., Ricci, L.A., & Slok, T. (2002). Capital account liberalization and economic performance: Survey and synthesis. IMF Working Papers 02/120. International Monetary Fund. Edwards, S. (2001). Capital mobility and economic performance: Are emerging economies in different. NBER. Working Papers 8076. National Bureau of Economic Research, Inc. Hansen, B.E. (1999). Threshold effects in non-dynamic panels: Estimation, testing, and inference. Journal of Econometrics, 93, 345–368. Islamaj, E. (2008). Financial integration and consumption smoothing: A survey. Kalemli-Ozcan, S., Sorensen, B.E., & Yosha, O. (2003). Risk sharing and industrial specialization: Regional and international evidence. American Economic Review, 93, 903–918. Kose, M.A., & Prasad, E.S. (2010). Emerging markets: Resilience and growth amid global turmoil. Washington DC: The Brookings Institution Press. Kose, M.A., Prasad, E., Rogoff, K., & Wei, S.J. (2009). Financial globalization: A reappraisal. IMF Staff Papers, 56, 8–62. Kose, M.A., Prasad, E.S., & Taylor, A.D. (2011). Thresholds in the process of international financial integration. Journal of International Money and Finance, 30, 147–179. Kose, M.A., Prasad, E., & Terrones, M. (2003). Financial integration and macroeconomic volatility. IMF. Working Papers 03/50. International Monetary Fund. Kose, M. A., Prasad, E. S., & Terrones, M. (2007). How does financial globalization effect international risk sharing: Patterns and channels. Economics. Kose, M.A., Prasad, E.S., & Terrones, M.E. (2009). Does financial globalization promote risk sharing? Journal of Development Economics, 89, 258–270. Lane, P.R., & Milesi-Ferretti, G.M. (2007). The external wealth of nations mark II: Revised and extended estimates of foreign assets and liabilities, 1970–2004. Journal of International Economics, 73, 223–250. Lewis, K.K. (1996). What can explain the apparent lack of international consumption risk sharing? Journal of Political Economy, 104, 267–297. Melitz, J., & Zumer, F. (1999). Interregional and international risk-sharing and lessons for emu. Carnegie-Rochester Conference Series on Public Policy, 51, 149–188. Milesi-Ferretti, G.M., & Razin, A. (1998). Current account reversals and currency crises: Empirical regularities. NBER Working Papers 6620. National Bureau of Economic Research, Inc. Neumann, R.M., Penl, R., & Tanku, A. (2009). Volatility of capital flows and financial liberalization: Do specific flows respond differently? International Review of Economics & Finance, 18, 488–501. Obstfeld, M. (1994). Are industrial-country consumption risks globally diversified? NBER Working Papers 4308. National Bureau of Economic Research, Inc. Pakko, M. (1998). Characterizing cross-country consumption correlations. The Review of Economics and Statistics, 80, 169–174. Quinn, D. (1997). The correlates of change in international financial regulation. The American Political Science Review, 91, 531–551. Razin, A., Sadka, E., 2001. FDI Flows and Domestic Investment: The Dominant Role of FDI. Mimeo. Tel-Aviv University. Suzuki, Y. (2014). Financial integration and consumption risk sharing and smoothing. International Review of Economics & Finance, 29, 585–598.

Please cite this article as: Malik, S., Financial-integration thresholds for consumption risk-sharing, International Review of Economics and Finance (2015), http://dx.doi.org/10.1016/j.iref.2015.01.004