“On the relationship between openness and macro economic volatility: A cross sectional study”

(1)

Faculty of economics and business

“On the relationship between openness

and macro economic volatility: A cross

sectional study”

Msc Thesis: International Economics and Business

Timme Spakman s2490145

Date: 15

th

of June, 2015

(2)

Figures, Tables, Equations, and Appendices

FIGURE 1: DYNAMICS OF ECONOMIC GROWTH VOLATILITY 1975-2005 2

FIGURE 2: TRADE AND FINANCIAL OPENNESS 1975-2005 2

FIGURE 3: THEORETICAL MODEL 10

TABLE 1: CORRELATION MATRIX 20

TABLE 2: DATA AND SOURCES 25

TABLE 3: DESCRIPTIVE STATISTICS BY COUNTRY GROUPS 31

TABLE 4: DESCRIPTIVE STATISTICS HIGH INFLATION COUNTRIES 32

TABLE 5: TIME DYNAMICS DEPENDENT VARIABLE 33

TABLE 6: PARTIAL ELASTICITIES FOR TRADE OPENNESS (GDP GROWTH VOL) 36

TABLE 7: MAIN REGRESSION RESULTS 37

TABLE 8: PARTIAL ELASTICITIES FOR CAPITAL A/C VOLATILITY (GDP GROWTH VOL) 38

TABLE 9: PARTIAL ELASTICITIES FOR TRADE % (INFLATION VOL) 39

TABLE 10: PARTIAL ELASTICITIES FOR CAPITAL A/C VOLATILITY (INFLATION VOL) 40 TABLE 11: PARTIAL ELASTICITIES FOR CAPITAL A/C VOLATILITY (CONS GROWTH VOL) 42

EQUATION 1: HERFINDAHL INDEX 24

EQUATION 2: TRADE OPENNESS AND INFLATION VOLATILITY 26

EQUATION 3: ECONOMIC GROWTH VOLATILITY 27

EQUATION 4: INFLATION VOLATILITY 28

EQUATION 5: CONSUMPTION GROWTH VOLATILITY 29

APPENDIX A: LIST OF COUNTRIES INCLUDED IN THE SAMPLE: 50

APPENDIX B: SUMMARY STATISTICS OF REGRESSION DATA. 51

APPENDIX C: CORRELATION MATRIX REGRESSION DATA 52

APPENDIX D: REGRESSION RESULTS 53

UNTIL

APPENDIX I: REGRSSIONS RESTULS 58

Acknowledgements:

(4)

Abstract

This paper studies the effects of economic openness on macro economic volatility. It measures macro economic volatility as the standard deviations of Economic growth, Inflation and Consumption Growth Volatility

The empirical analysis of this paper is build upon un-balanced panel data regressions, incorporating a sample over 55 countries for the period between 1982 and 2009. The analysis yielded significant evidence for the hypotheses: That (foreign) capital account volatility works destabilising on economic growth for countries with low financial deepness; trade openness has a stabilising (destabilising) effect on inflation for countries with high (low) export concentration; and capital account openness has a stabilising effect on

(5)

1

I. Introduction

One feature of the world today, or at least until the ’07 ’08 great recession is, that it appears to be more stable than a couple of decades ago. Cecchetti, Lagunes and Krause (2005) note that both, average output growth volatility as well as inflation volatility declined significantly in the period between the early eighties and 2005. From the data used in this study (volatility measured as the average1 standard deviations of real economic growth, inflation, and consumption growth), this downward trend seems only pronounced for high-income countries2. For high-income countries, volatility peaked again in more recent years. For low, lower-middle, and upper-middle income countries the situation since the early eighties seems almost unchanged.

To understand the possible causes and consequences of macro-economic volatility, as well as the implications of its dynamics, one should take notice of its essence. Calderón and Schmidt-Hebbel (2008) define the underlying concept of macro economic volatility as a country’s exposure and vulnerability to shocks. Exposure to shocks could be a country’s exposure to global price shocks, climate conditions, or sector dependence. Where the degree to which these shocks are amplified or mitigated is a reflection of a country’s vulnerability to economic shocks.

Most literature, reviewed for this paper, proxy economic volatility using the standard deviation of economic growth, exampli gratia: Bekaert, Harvey, and Lundblad (2006); Celderón and Schmidt-Hebbel (2008); Cavallo, Gregorio, and Loayza (2008); Klomp and de Haan (2009); and Mallick (2014). However, when considering the definition of economic volatility as formulated by Calderón and Schmidt-Hebbel (2008), one easily arrives at the conclusion

1

Average standard deviation computed using a non rolling-window method over four periods of seven years. The time frame for each period is respectively 1982-1988, 1989-1995, 1996-2002, and 2003-2009.

2

(6)

2 that macro economic volatility is a wider concept than just economic growth variability. Some studies with a scope extended beyond economic growth volatility, are Kose, Prasad, and Terrones (2003) and Bekaert, et al (2006) who also consider income and consumption growth volatility, and Bowdler and Malik (2005) who focus specifically on inflation volatility.

Calderón and Hebbel-Schmidt (2008) compare a larger sample and come to similar findings as Cecchetti et al (2005). Figure 1 clearly shows a declining trend of GDP growth volatility for all groups of countries.

Figure 1: Dynamics of Economic Growth Volatility 1975-2005

Where: figure has been sourced from Calderón and Hebbel-Schmidt (2008, Appendices

figure one)

During this same period, openness to trade and foreign finance increased rapidly (Figure 2). This apparent correlation has been the motivation for the majority of the studies looking into openness and economic volatility.

Figure 2: Trade and Financial Openness 1975-2005

Where: figure has been sourced from Calderón and Hebbel-Schmidt (2008, Appendices

figure one)

(7)

3 openness expose countries to global price shocks and volatile capital flows that harm economic stability (Easterly, Islam and Stiglitz 2000; Caballero and Krishnamurthy 2001), while others ague, inline with the seemingly correlating time trends (Figure 1 and Figure 2), that enhanced possibilities for international risk sharing and consumption smoothing cause openness to have a stabilising effect on economic performance (Kose et al 2003 and Dynan, Elmendorf and Sichel 2005).

The aim of this paper is to study the potential relationship between openness and macro-economic volatility and to contribute to the current limited amount of literature, by: incorporating more recent data and reviewing the consistency of the effects of openness on three different forms of macro economic volatility, namely: economic growth volatility, inflation volatility and

consumption growth volatility.

The causes of macro economic instability are relevant to study as economic volatility, potentially, affects growth, consumption, and economic development. Although, also these relationships are still strongly debated, a large number of studies find a negative relationship between economic volatility (as measured by GDP volatility) and economic growth (Ramey and Ramey 1995; Fatás 2002; and Hnatkovska and Loayza 2005). In addition Mobarak (2005) argues that economic volatility in particular affects those with the smalest incomes, as they are more sensitive to fluctuations to (real) income.

(8)

4

II. Statement of Research Question

This introductory chapter offers the rational for the research questions. Section 2.1 touches upon the relevance of the dependent variables and section 2.2 gives an brief overview of the literature predicting economic (in)stability. This chapter concludes with the introduction of the research questions in section 2.3.

2.1. The Relevance of Macro Economic Stability

This section addresses relevance of the dependent variables Subsection 2.1.1 discusses economic growth volatility and subsection 2.1.2 Addresses the relevance of inflation volatility.

Not much literature could be found on the economic consequences of consumption growth volatility. However, this variable could be useful to consider the consistency of the effects of openness on economic volatility. In particular because consumption growth volatility is a driving force behind economic growth volatility. (Blanchard and Simon 2001)

Economic growth variability 2.1.1

The debate around the relationship between growth volatility and economic growth gained pace after the empirical work of Ramey and Ramey (1995) who conducted a cross-sectional study covering the period between 1960 and 1985. They found a significant negative relationship between the standard deviation of economic growth and economic growth rates, which is confirmed by others to have, at least, persisted into the 1990’s (Fatás 2002 and Hnatkovska and Loayza 2005).

(9)

5 output growth volatility and economic growth, they show that on the sectoral level, growth and variability in sector growth are positively related. They explain this by the process of creative destruction. In addition some studies show that different causes of output growth volatility have different impact on economic growth (Kose et al 2005; Fatás and Mihov 2003). Kose et al (2005) finds that the relationship between volatility and economic growth changes with trade and financial liberalisation. While they provide evidence that in general volatility of economic output increases with integration into the global economy, they provide evidence that trade openness mitigates the negative relation between fluctuations in output and growth. They conclude that countries that open-up to trade might experience stronger fluctuations in growth rates, however, their economic growth performance might be less affected by this increased volatility. While Kose et al (2005) find that increased openness to international trade mitigates the relationship between output growth variability and economic growth; they also provide evidence that financial openness strengthens the negative relationship between output volatility and economic growth.

In addition, as mentioned in the introduction, Mobarak (2005) finds that output volatility hurts consumption, in particular for the smallest incomes.

Although the relationship between economic growth-volatility and growth is far from straightforward, it could be concluded that it is a relevant factor explaining some part of economic growth.

Inflation Volatility 2.1.2

(10)

6 correlation does not imply causation and exogenous factors might influence both.

Emara (2012), Judson and Orphanides (1999), and Al-Marhubi (1998) find that inflation volatility has a negative impact on economic growth. Temple (2002) argues that the cost of inflation variability is potentially higher in more integrated economies. The harmful effects of inflation volatility where discussed a long time ago by Friedman (1977) He argues that higher volatility in inflation rates reduces the optimal length of market transactions and makes it harder to extract the right price from market indices. Huizinga (1993) argues that increased uncertainty on future inflation rates will encourage firms to delay investments unless the net present value (NPV) of projects is large enough to make up for the increased uncertainty. He studies the effects of inflation uncertainty on the US manufacturing industry using a 35-year panel regression model and finds positive evidence for his hypothesis. Bowdler and Malik (2005) argue that inflation volatility is more costly in an open economy because it deteriorates the competitive position of domestic suppliers, and that inflation stability might promote income growth.

2.2. Determinants of Macro Economic Volatility

Cecchetti et al (2005) sum five major (possible) explanations for the decrease in economic growth volatility and inflation volatility, these, complemented by some of the literature reviewed for this study, are:

 First: improved inventory management practices could have played a mitigating role for demand shocks (McConnel, Kahn and Perez-Quiros 2002 and McConnel and Kahn 2005).

 Second: more sound and disciplined monetary policy (Cecchetti and Krause 2002).

(11)

7

 Fourth Increased commercial openness and more intensive trade flows (Bowdler and Malik [inflation volatility] 2005; Cavallo 2006; and Calderón and Schmidt-Hebbel 2008).

 Fifth: Luck in the form of smaller business cycle shocks and a relatively stable period, which is rather caused by an arbitrary phenomenon than by dynamics in economic factors. (Ahmed, Levin and Wilson 2002 and Stock and Watson 2002)

In addition to this, Mobarak (2005) and Klom and de Haan (2009) find that democracy has a stabilising effect on the standard deviation of economic growth. Furthermore, Klom and de Haan find in their analysis, incorporating over 100 countries for an extensive time period, from 1960 to 2005, some evidence for a relationship between political instability and policy uncertainty and output growth volatility.

Although, aforementioned has been suggested that trade and financial openness play potentially an important role in explaining macro economic volatility, the literature is ambiguous on these relationships is rather ambiguous. Calderón and Hebbel-Schmidt (2008 p.2) state:

“Empirically, there is no consensus on the effects of trade and financial openness on growth volatility”

and Kose et al (2003 p.123) who state the following on the relationship between economic openness and economic volatility:

“Existing studies have been unable to document a clear empirical link...”

(12)

8

Trade openness and Economic volatility 2.2.1

Giovanni and Levchenko (2009) find that trade openness enhances growth volatility, but also state that currently there is no consensus on the nature of the relationship between trade openness or integration and the volatility of economic growth. Kose Prasad and Terrones (2003) state that the theory between trade openness and growth volatility is ambiguous, and the relationship differs with the degree of specialisation in intra versus inter industry trade. Haddad Lim Pancaro and Saborowski (2013) provide evidence that the relationship between trade openness and growth volatility overall is negative but that the degree of diversification in export moderates this relationship. Although growth volatility is widely considered in the relation to trade liberalisation, little literature is available on the effects of trade openness on inflation volatility and consumption growth volatility.

Financial Openness and Economic Volatility 2.2.2

(13)

9 development to mitigate macro-economic volatility. However Easterly et al (200) find no significant effect for financial openness or hot-money flows on macro economic volatility.

2.3. Chapter Summary and Research Questions

In conclusion, economic volatility in the different forms discussed in this chapter (Economic, Consumption growth volatility and Inflation Volatility) are highly relevant variables due to their (potential) influence macro economic performance. Although the influence of economic openness on the different forms of economic volatility is far from straightforward, there are a number of studies that find significant effects. In addition, the relationship is also relatively unexplored. On the basis of the discussion in this chapter I have formulated the following research questions:

RQ 1: What are the effects of economic openness on macro-economic

volatility?

RQ 2: Are these effects consistent over the different aspects of

economic openness, id est: trade openness and financial openness.

RQ 3: Are these effects consistent over the different form of economic

(14)

10

III. Theoretical Model

This chapter explains the theoretical model and introduces the different hypotheses.

The theoretical model of this paper is graphically depicted below. Although this model is roughly the same for all three dependent variables, the line of reasoning sometimes slightly differs. The rational arguments behind the relations shown in Figure 3 are explained in this chapter for each dependent variable. Section 3.1 addresses the relationship between openness and

economic growth volatility, section 3.2 gives an overview of the relationship

between openness and inflation volatility, and section 3.3 addresses the relationship between openness and consumption growth volatility.

Figure 3: Theoretical Model

Where: Solid lines indicate hypothesized positive relationships, and Dashed lines relationships

(15)

11 3.1. Openness and Economic Growth Volatility

Trade Openness and Economic Growth Volatility 3.1.1

Trade integration has, as mentioned in section 2.2.1, not a clear cut effect on the volatility of economic growth rates. The work of Haddad et al (2013) suggests that the ambiguous results in the literature might stem from the idea that the relationship is moderated by export diversification. Haddad et al (2013, 766) conclude that:

“Irrespective whether the effect of trade openness on output volatility is positive or negative on average, openness lowers output volatility in sufficiently diversified economies, while it increases volatility in those with more concentrated

export baskets.”

This is based on the theoretical argument that openness to international trade increases a country’s exposure to idiosyncratic shocks from abroad, which fuel economic volatility (Haddad et al 2013). However, openness to trade comes potentially with the positive externality of international risk sharing. When two countries engage in international trade and have asynchronous business cycles, the reduction in domestic demand due to a bust in one country is mitigated by the increased foreign demand from the booming country. This example assumes that relative prices in both countries are equal or at least in favour to stimulate foreign demand for the exports of the country that experiences the downturn. (de Grauwe, 2012)

(16)

12 level of diversification in the export basket makes a trading nation more vulnerable for foreign demand shocks and stronger Terms of Trade swings. Thus sufficient export diversification3 cancels out idiosyncratic shocks such that the risk sharing properties of international trade prevail (Haddad et al 2013; and Cavallo et al 2006). Calderón and Schmidt-Hebbel conduct a panel data analysis for 82 to countries over the period from 1975 to 2005 and find that trade has a stabilising effect on growth but works destabilising in countries with high concentration in exports.

Therefore I hypothesise that:

H1.1 Trade openness has no independent effect on economic growth volatility.

H1.2 The relationship between trade concentration and economic growth volatility is negative for countries with a high degree of export diversification and positive for countries with high concentrated exports.

Financial Openness and Economic Growth Volatility 3.1.2

Economic theory offers little explanation on how financial openness should affect economic growth variability. Kose et al (2005) argue that financial openness might reduce economic shocks by the provision of international capital flows for investment in a more diversified export basket. However, this could also work in the opposite direction when capital is utilised for deepening a country’s specialisation in exports.

Potentially financial openness increases international investment opportunities and facilitates a wider assortment of financial instruments. This allows for more advanced (international) risk-sharing schemes, and might therefore contribute to more stable output (Calderón and Schmidth-Hebbel 2008).

3

(17)

13 Klein and Olivei (2008) argue that the relaxation of capital account restrictions potentially contributes to the development of the financial system due to exposure to international competition and the introduction of international standards. However, this is conditional to the quality of institutions. Roderik (1999) and Klein (2005) argue that in order to profit from capital account openness, the presence of decent institutions are important is key. A decent rule of law and protection of property rights makes it more attractive for financial firms to expend. In addition, as history has thought us, not having the institutional framework in place to stabilise capital flows can be dangerous As Roderik (1999 p. 30) states:

“ Openness to international capital flows can be especially dangerous if the appropriate controls, regulatory apparatus and macroeconomic frameworks are not in

place”

Financial openness increases exposure to volatile capital flows (Kose et al 2003). In the simple Keynesian model, positive shocks in the money supply raise inflation and output. Where an inflow (outflow) of foreign capital might act as an increase (decrease) in the money supply, volatility in capital flows might induce volatility in output and inflation (Unless these flows are sterilised by the central bank). Forbes and Warnock (2012 p.235) state that capital volatility has the potential to

“…amplifying the economic cycle and increasing financial system vulnerabilities and overall macro-economic instability“

(18)

14 (2002) do not find any significant effects of sudden stops in capital flows on the volatility of output nor consumption.

O’Donnel (2001; cited from Kose et al 2003 original paper not

published) studies the effects of international financial integration on

economic growth variability over an extensive period of time. He finds that higher financial integration induces higher (lower) output volatility for non-OECD (OECD) countries. This might be consistent with the model of Caballero and Krishnamurthy (2001) and Aghion et al (1999) Roderik (1999) and Klein (2005) since non-OECD countries are less likely to have strong developed financial sectors and institutions. Calderon and Schmidth-Hebbel (2008) find that financial openness and integration overall have a destabilising effect on economic growth.

Before formulating the hypotheses regarding financial integration, I would like to conclude with the notion that overall the relationship between capital account openness seems to be off consensus. As Kose et al (2003) p.123 note that:

“Unlike the rich empirical literature focusing on the impact on financial openness on economic growth, there are only a limited number of studies analysing

the links between openness and macro economic volatility.” And “Existing studies have been unable to document a clear empirical link…”

Based on the studies reviewed in this section I have drawn the following hypotheses:

H1.3 Capital account volatility causes stronger economic growth volatility H1.4 The relationship between capital account volatility and economic

growth volatility is more pronounced for countries with weaker developed financial sectors and financial deepness.

Due to more advanced possibilities for international risk sharing I hypothesise that:

(19)

15 3.2. Openness and Inflation Volatility

Before addressing consumption growth volatility I first discuss the relationship between openness and inflation volatility. This is because inflation volatility is, potentially, a mediating variable in the transmission from openness to consumption growth volatility.

As mentioned in section 2.1.2 inflation volatility is correlated with economic growth variability (Blanchard and Simon 2001). They argue that this does not necessarily imply causation; although this correlation still might imply indirect causation, I hypothesise that forces related to openness could affect both.

Resulting from the lack of literature on the relationship between openness and inflation stability, the theoretical arguments in this section are almost exclusively backed by the paper of Bowdler and Malik (2005). They conduct a cross-country study over the period of 1961 to 2000. They provide evidence that openness may contribute to less inflation volatility.

Razin et al (2003) developed a theoretical model that predicts that openness enhances volatility due to terms of trade shocks. Bowdler and Malik (2005) argue that it is a common view that trade openness increases exposure to global price shocks. This is however conditional on the level of export concentration. This is related to the discussion in section 3.1.1; where countries are well diversified in their exports, domestic price shocks could be mitigated due to the law of large numbers (natural hedging).

In developing countries, inflation instability finds partially its roots in price fluctuations of agricultural sector goods, driven by fluctuations in climate conditions. More open economies could in such case benefit from more price stability in the agricultural sector as surpluses (shortages) in supply could be exported (imported) (Aron and Muelbauer 2000).

(20)

16 governments and central banks in open economies might be keener to pursue more disciplined monetary policy and inflation targeting. Therefore openness might have a stabilising effect on inflation volatility (Bowdler and Malik 2005).

Since for countries, with a well-diversified export basket, terms of trade shocks are mitigated by natural hedging (Bowdler and Malik 2005), and countries that are more open to trade are potentially keener to pursue more decent monetary policy (Muellbauer 2000), I hypothesise that trade openness has a stabilising effect on inflation volatility given that exports are sufficiently diversified.

H2.1 Trade openness has no independent effect inflation volatility.

H2.2 The relationship trade openness and inflation volatility is negative for countries with a high degree of export diversification and positive for countries with highly concentrated exports.

In addition to trade openness, financial openness also might have a direct effect on inflation volatility. As Bowdler and Malik (2005) argue that in particular foreign credit and capital reversal are related to recessions and inflation volatility. This is, as mentioned in subsection 3.1.2, mitigated by strong financial sectors.

H2.3 Capital account volatility causes stronger inflation volatility

H2.4 The relationship between capital account volatility and inflation volatility is more pronounced for countries with weaker developed financial sectors and financial deepness.

Inflation volatility might, in an open economy, induce foreign investors to repeal their capital. Therefore, Financial openness might also offer incentives for authorities to be more disciplined in monetary policy. Therefore I also hypothesise that:

(21)

17 3.3. Openness, Consumption Growth Volatility

The relationship between openness and consumption growth volatility is addressed in the studies of Baekert et al (2006), Kose et al (2003), and Baxter and Crucini (1995). The first two studies are mainly focused on the empirics. Although they articulate some arguments, they do not provide a comprehensive theory why openness should affect consumption growth volatility. Baxter and Crucini (1995) formulate a theory centralised on international risk sharing.

Based on the permanent income hypothesis of Friedman (1957), I offer some theoretical arguments why consumption growth volatility might be induced or stabilised by economic openness.

Section 3.3.1 discusses the relationship between trade openness and consumption growth volatility and section 3.3.2 addresses the relationship between financial openness and consumption growth volatility.

Trade Openness and Consumption Growth Volatility 3.3.1

Consumption Growth volatility is affected by economic openness mainly via the terms of trade (Kose et al 2003).

Inflation volatility might have a role in channelling the degree of trade openness to consumption growth volatility. The rational behind this argument is that terms of trade volatility potentially induces inflation volatility (Bowdler and Malik 2005). Logically, instable inflation affects the stability of real income. Increased price uncertainty induced by terms of trade shocks might offer incentives to households to increase savings in order to smooth consumption more efficiently. This, then, would induce short episodes of volatility in consumption growth when households adjust their savingsrate to the new level of price and real income uncertainty.

(22)

18 exposure to large terms of trade shocks that have the potential to change the level of price uncertainty. This is, as explained in section 3.2, dependent on the level of export concentration.

H3.1 Trade openness has no independent effect on consumption growth volatility.

H3.2 The relationship between trade concentration and consumption growth volatility is negative for countries with a high degree of export diversification and positive for countries with high concentrated exports.

Financial Openness and Consumption Growth Volatility 3.3.2

According to Kose et al (2003) and Baxter and Crucini (1995), financial openness increases the potential for international risk sharing and consumption smoothing. An open capital account allows for more divers (international) investment opportunities, and stimulates further financial integrations. This leads to a more divers assortment of financial instrument and, hence, more advanced possibilities for (international) cross sectional risk sharing. Baekert et al (2006) use this argument to hypothesise that consumption growth volatility decreases with financial integration. Although Baekert et al (2006) find that financial integration is related to lower consumption growth volatility, Kose et al (2003) find a destabilising effect on consumption growth volatility.

My reception of this argument on international risk sharing is rather critical. Assuming that for the average person, income per capita is high enough to save. Households are able to (intertemporally) smooth consumption, simply by saving and consuming savings. Following this line of reasoning, households do not necessarily have to borrow internationally to smooth consumption over time.

(23)

19 and banks. However, considering the permanent income hypothesis (Friedman 1957), this still does not imply more stable consumption. Enhanced opportunities for international risk sharing might raise the implicit value of (lifetime) savings and pensions, which would only imply a lower savings rate and higher consumption levels and not necessarily more stable consumption.

A primary critique on the permanent income hypothesis is that it is based on the assumption of rationality while in reality people do often act irrational. The relaxation of this assumption provides the space for a potential transmission mechanism from financial openness to consumption growth volatility.

Mass hysteria driven by economic booms and busts might induce consumption growth volatility. This is due to adjustments in consumption levels based on perceived changes in permanent income rather than actual changes.

In line with this argument, capital account openness might have a destabilising effect on consumption growth. Capital account openness increases exposure to volatile capital flows, which have the potential to magnify boom and bust cycles (Warnock 2012). This might destabilise consumption growth, as larger boom-bust cycles, cause households to adjust spending behaviour more pronounced. These fluctuations in consumption growth are in such cases the symptoms of adjustment after a change in (perceived) long-term income.

H3.3 Capital account volatility causes stronger economic consumption growth volatility

(24)

20 3.4. Quality of Institutions

As mentioned in section 2.2, institutions have a direct effect on macro economic volatility. Mobarak (2005) and Klom and de Haan (2009) show that in particular democracy mitigates economic volatility.

Apart from the moderating effect on the relationship with capital account openness and economic volatility, the quality of the regulatory apparatus is expected to have a stabilising effect on all three forms of macro economic volatility. This is partly due to the logic that a more stable and better functioning legal framework gives authorities better tools and incentives to engage in stabilising polices. In addition, a stronger rule of law is negatively correlated with socio political instability, which, has been shown to have a destabilising effect on economic performance Klom and de Haan (2009). As shown in table 1, Rule of Law and the absence of corruption are both strongly correlated with capital account openness. Therefore it is important to control for institutional quality to mitigate potential omitted variable bias.

Table 1: Correlation Matrix

Cap Openness Trade Openness Regulatory Quality Rule of Law Corruption

Cap Openness 1

Trade Openness 0.16 1

Regulatory Quality 0.61 0.28 1

Rule of Law 0.54 0.29 0.91 1

No-Corruption 0.53 0.28 0.87 0.97 1

Where: Capital account Openness is provided by Chinn and Ito (2008), Trade Openness is

measured as trade flows as a percentage of GDP (World Bank Data), and Regulatory Quality, Rule of

(25)

21 3.5. Chapter Summary and Concluding Remarks

This chapter introduced the central theory and the corresponding hypotheses. The main relationships addressed in this chapter are:

 Trade openness has a stabilising effect on economic volatility. The exact nature of the relationship differs per dependent variable. However, the relationship is hypothesised, for all three forms of economic volatility (Economic growth-, Inflation-, and Consumption Growth-

volatility), to be negative when exports are sufficiently

diversified.

 Capital account openness mitigates economic volatility in the case of Economic growth-, and Inflation Volatility.

 Capital account openness enhances economic volatility, for all three forms (Economic growth-, Inflation-, and Consumption Growth- volatility). This is transmitted via volatile capital flows. This relation is expected to be more pronounced for countries with a weaker financial sector and less financial deepness.

(26)

22

IV. Data and Methodology

This section gives an outline of the data and explains the regression models used to test the hypotheses formulated in chapter III. Section 4.1 gives an overview and justification of the data and its sources, Section 4.2 addresses the methodology, and Section 4.3 elaborates on the data set using descriptive statistics.

4.1. Data

A general overview of the data and data sources are shown in

Table 2. In this subsection I discuss the general characteristics of the data and its limitations.

Growth volatility is computed as the standard deviation of real per capita GDP growth over a period of seven years. Where real GDP per capita growth is computed as the percentage change in real GDP divided by population size. Where: real GDP is at 2005 constant USD. Both population size and Real GDP are sourced from the PENN world tables v. 8.1 (Feenstra Inklaar, and Timmer, 2015). With exception of Nicaragua and Hongkong, which are not included in the Penn world data set and sourced from the World Bank Real (GDP per capita at constant 2005 USD).

Consumption growth volatility has been computed from extracted from the World Bank databases. The measure used is household expenditure as a % of GDP (formerly private consumption). The volatility measure is computed as the standard deviation over periods of seven years of percentual growth rates of consumption.

To proxy average inflation and inflation volatility I use CPI inflation. Inflation volatility is calculated as the standard deviation of inflation over periods of seven years.

(27)

23 gross capital flows for 58 countries over the period from 1980 to 2009. They use four categories: “Surges”, “Stops”, “Retrenchment”, and “Flight, which respectively mean, sharp increases in foreign capital inflows, sharp decreases in foreign capital inflows, sharp decreases in domestic capital outflows, and sharp increases in domestic capital outflows. The dummy variables indicate for each category when one of the flows deviates more than one standard deviation from its historical mean. Where the historical average is computed using rolling windows over a period of five years.

The first constructed measures used in the analysis to proxy capital account volatility, is a proxy of total volatility in capital flows. It has been computed as the sum of the binary values for all four categories for periods of seven years. This measure reflects the number of quarters that capital flows (independent form which category) deviated more that one standard deviation from its historical average. The second measure has the same structure as the first, but only includes foreign capital flows (Surges and Stops) as sub components.

(28)

24 To quantify capital account openness I used the Chinn and Ito index (2012). This is a constructed measure based on binary variables codifying capital account restrictions initially reported in the IMF’s Annual Report on Exchange arrangements and Exchange Restrictions (AREAER).

Credit as percentage of GDP is the seven year average of domestic credit to the private sector as percentage of GDP.

The primary measure of export concentration is the IMF export concentration index. This measure of export concentration is two-dimensional and includes three different variables. In this study the Theil index for export concentration along the intensive margin is used to proxy export concentration. It is important to note that the intensive and extensive margins have not the exact same definitions as the aforementioned dimensions of export diversification (Chapter III), id est: number of products and number of trade partners. Concentration in the intensive margin reflects the inequality of exports over export lines; where increasing concentration along the extensive margin reflects a smaller number of (active) export lines. Where export lines could be either different products exported to the same country or the same products exports exported to different countries.

To test the consistency of the effects I also use a secondary measure of export concentration, which is the seven year average of the Herfidahl index calculated from the UN-Comtrade statistics, STIC revision 2, on a 3-digit scale. The Herfidahl index is a concentration measure computed as the sum of the squared relative export shares (Equation 1).

Equation 1: Herfindahl Index

∑ (

∑ )

Where: the herfindahl index is computed for a specific country for a specific time, are the USD export

(29)

25 To measure the quality of institutions I use two proxies. The first is the rule of law index provided by the cline centre for democracy. This rule of law measure is based on a common factor analysis (Nardulli Peyton and Bajjalieh 2013). I particular choose this Rule of Law index because it is available for an extensive period of time. As a control variable for democracy I use the political rights measure offered by freedomhouse.org.

Table 2: Data and Sources

VARIABLES SOURCES NOTES

Real GDP per capita & Output Volatility

Penn World Tables 8.1 (Feenstra et al, 2015): computed from indicators rgdpna and pop. Output volatility has been computed as the standard deviation over periods of 7 years of the annual percentage change in Real GDP per capita

Hong Kong and Nicaragua are sourced from the World Bank (NY.GDP.PCAP.KD).

Consumption Growth Volatility

Sum of Household and public consumption World Bank Data: House hold expenditure as % of GDP (NE.CON.PETC.ZS)

Computed as the standard deviation over periods of 7 years of the annual percentage change in Household expenditure

Inflation & Inflation Volatility

World Bank Data (FP.CPI.TOTL.ZG) Inflation volatility is proxied as the standard deviation over periods of 7 years.

With exception of Argentina, Bangladesh, Chile, Germany, UK, Nicaragua, and Venezuela; which are sourced from IMF world Economic Outlook via Index Mundi (www.indexmundi.com).

Capital Account volatility

Sourced from: Forbes and Warnock (2012) online Appendix

Data for Belgium and Luxembourg are included as one country. Double entries and type errors make accuracy for the countries South Korea and Turkey questionable.

Capital Account Openness

Chinn and Ito Index for capital account openness updated to 2012 (Chinn and Ito, 2006)

Credit % of GDP World Bank: Domestic credit to private sector as % of GDP (FS.AST.PRVT.GD.ZS)

Export Concentration

Theil indices of concentration (IMF 2014) and the Herfidahl index computed from 3 digit STIC rev. 2 UN COMTRADE stats.

As primary proxy, the intensive margin Concentration Measure from the IMF data set has been used.

Political rights Freedomhouse.org Adjusted the ranking from 1 highest (7 lowest) to 7 Highest (1 lowest).

Rule of Law Cline Center for Democracy (Nardulli, Peyton, and Bajjalieh 2013)

(30)

26

Where all data is computed as averages or standard deviations over 7 year periods for the time frame from 1980 to 2009. All modifications are included in the table, and a list of countries could be found in Appendix A

4.2. Methodology

This section addresses the structure of the regression models. I first explain the general characteristics of the regression equations, and go into the specifics for each Hypothesis in the remainder of this section. The general structure of the model is shown in the equation below.

Equation 2: Trade Openness and Inflation Volatility

( ) ( ∑ [ _] ) (∑ ) (∑ [ ] ) (∑ ) (∑( ) ) (∑ ) (∑ ) (∑( ) ) (∑ )

Where: The dependent variable is economic volatility for country over period , which is a time frame

of 7 years ( ). denote the quarters for each period , at which is the interval at, which, the binary values for the four categories of volatile capital flows are measured.

(31)

27 advantage of no overlapping data, which would have been the case when measuring volatility with rolling window standard deviations. All variables are therefore either standard deviations or averages over the four seven-year periods.

Most variables are positively skewed. To normalise the data I have chosen for a regression equation with a log-log / log-linear functional form. Explanatory variables that are negatively skewed have not been normalised to maintain the ease of interpretation.

A welcome externality of log-log regression equations is that coefficients are interpretable as elasticity’s. This makes it easier to compare coefficients for different variables. In order to control for unobserved country differences I apply all regression models with country fixed effects.

Economic Growth Volatility 4.2.1

To test the hypotheses H1, I use the natural logarithm of the standard deviation of GDP growth (equation 3), as dependent variable. Equation 3: Economic Growth Volatility

(√∑ ( ̅̅̅̅̅̅̅̅) )

Where: is the GDP per capita growth rate for country at time , where Period equals the

timeframe of 7 years.

Trade openness (average trade % of GDP) is hypothesised (in line with H1.1 (Trade has no independent effect on economic growth

volatility) not to have a significant effect when its interaction effect with

export concentration is excluded from the analysis. However, when average concentration of exports is added as a factor variable with trade openness, 1 is expected to be negative and 3 is expected to be positive (in line with H1.2 The relationship between trade

(32)

28

with a high degree of export diversification and positive for countries with high concentrated exports).

The effects related to financial openness are included in the second half of Equation 2; the coefficient ( 4) of capital account volatility is expected to be positive, in line with hypothesis H1.3 (Capital

account volatility causes stronger economic growth volatility). To test

the dependence of this relationship with financial development I have added an interaction variable with credit to the private sector as a percentage of GDP. The coefficient of this interaction effect ( 6) is hypothesised to be negative in line with H1.4 (The relationship between

capital account volatility and economic growth volatility is more pronounced for countries with weaker developed financial sectors and financial deepness).

The coefficient of capital account openness (

) is hypothesised to have a positive sign (H1.5 Capital account openness

reduces economic growth volatility). Inflation Volatility

4.2.2

To test hypothesis H2, I use the natural logarithm of the standard deviation in consumer price index inflation as dependent variable (equation 4).

Equation 4: Inflation Volatility

(√∑ ( ̅̅̅̅̅̅̅) )

Where: is the household consumption growth rate for country at time , where Period equals

the timeframe of 7 years.

(33)

29 concentration ( 3) should have, respectively, a negative and positive coefficient, in line with H2.1 (Trade has no independent effect on

inflation volatility) and H2.2 (The relationship between trade concentration and inflation volatility is negative for countries with a high degree of export diversification and positive for countries with high concentrated exports).

In correspondence with H2.3 (Capital account volatility causes

stronger inflation volatility), the coefficient for volatile capital flows ( 4)

is expected to have a positive sign. The coefficient of the interaction effect with credit as a percentage of GDP ( 6), is hypothesised to be negative in line with H2.4 (The relationship between capital account

volatility and inflation stability is more pronounced for countries with weaker developed financial sectors and financial deepness). The

coefficient for capital account openness ( 7) is hypothesised to have a negative sign (H2.4 Capital account openness reduces inflation

volatility).

Consumption Growth Volatility 4.2.3

To test H3, I use the logarithm of the standard deviation in annual consumption growth as the dependent variable (Equation 5).

Equation 5: Consumption Growth Volatility

(√∑ ( ̅̅̅̅̅̅̅̅) )

Where: is the household consumption growth rate for country at time , where Period equals

the timeframe of 7 years.

In line with hypothesis H3.1 (Trade Integration has no

independent effect on consumption growth volatility), I expect that the

(34)

30 the regression. However, when including the interaction effect, 1 is expected to be negative and the coefficient of the interaction effect ( 3) is expected to be positive (H3.2 The relationship between trade

concentration and consumption growth volatility is negative for countries with a high degree of export diversification and positive for countries with high concentrated exports).

I expect that volatile capital flows have a positive coefficient (in line with H3.4 Capital account volatility causes stronger economic

consumption growth volatility). The coefficient for the interaction effect

with credit as a percentage of GDP ( 6), is hypothesised to be negative (H2.4 The relationship between capital account volatility and

consumption growth volatility is more pronounced for countries with weaker developed financial sectors and financial deepness).

(35)

31 4.3. Descriptive statistics

Cross-country differences in Real GDP per Capita, inflation, and the dependent variables are shown in Table 3.

Table 3: Descriptive Statistics by country groups

Income Group Descriptive Statistics Real GDP Per Capita GDP Growth Volatility % Consumption Growth Volatility % Inflation % Inflation Volatility % Low Income Countries Mean 2277.4 1.96 2.47 9.65 4.91 Median 2608.6 1.67 1.94 9.04 3.25 Standard Deviation 897.1 1.41 1.68 3.93 4.76 Minimum 844.9 0.48 0.68 2.92 1.03 Maximum 3633.9 6.32 8.62 18.13 17.55 N 22 22 22 22 22 Low Middle Income Countries Mean 7159.7 2.72 3.07 71.43 43.42 Median 7441.5 2.52 2.90 8.28 4.26 Standard Deviation 2597.2 1.32 2.06 238.25 158.16 Minimum 1180.5 0.62 0.64 2.86 0.90 Maximum 12152.2 5.20 9.52 1404.68 929.51 N 36 36 36 36 36 High Middle Income Countries Mean 14534.5 3.22 2.39 29.65 35.96 Median 14326.2 2.43 1.89 5.86 3.00 Standard Deviation 4109.8 2.32 1.75 125.67 190.07 Minimum 8072.5 0.78 0.50 2.06 0.58 Maximum 23052.1 9.97 9.61 802.29 1216.68 N 41 41 41 41 41 High Income Countries Mean 29959.8 2.12 1.76 3.45 1.61 Median 27552.8 1.89 1.39 2.49 0.96 Standard Deviation 8563.4 0.95 1.16 4.32 2.43 Minimum 18785.4 0.55 0.38 -0.08 0.31 Maximum 61974.4 4.63 6.71 37.36 21.51 N 83 83 83 83 83 Full Sample Mean 18628.7 2.47 2.24 23.55 18.02 Median 18488.2 2.07 1.77 4.58 1.99 Standard Deviation. 12636.7 1.55 1.63 123.08 114.77 Minimum 844.9 0.48 0.38 -0.08 0.31 Maximum 61974.4 9.97 9.61 1404.68 1216.68 N 182 182 182 182 182

Where: Countries have been classified using the GNI country classification definitions of the World Bank. GNI has been sourced from the UN databases.

As expected economic volatility is on average lower in high-income countries relative to middle-high-income countries.

(36)

32 of Hyperinflation: Argentina and Brazil for the period 1982 to 1995. As a robustness check I also ran the regressions excluding cases of hyperinflation. I define the threshold inflation rate as an average annual increase of 60% in CPI over the periods of seven years. Although the common definition of hyperinflation is a monthly increase in CPI of over 50%, I consider a lower threshold of 60% per annum to exclude cases of very high inflation and very short episodes of hyperinflation. In total eight observations where removed from the data set, which are: Argentina and Brazil for the periods from 1982 to 1988 and 1989 to 1995, Israel and Mexico for the period 1982 to 1988, and Turkey for the period from 1989 to 1995 and 1996 to 2002. The contrast in the data set including and excluding these cases of high inflation is shown in table 4.

Table 4: Descriptive Statistics High Inflation Countries

Income Group Descriptive Statistics Real GDP Per Capita GDP Growth Volatility % Consumption Growth Volatility % Inflation % Inflation Volatility % Excluding Hyperinflationary Countries Mean 19064.34 2.40 2.16 6.47 3.23 Median 19320.42 2.02 1.73 4.41 1.89 Standard Deviation 12741.79 1.52 1.59 7.07 4.82 Minimum 844.88 0.48 0.38 -0.08 0.31 Maximum 61974.36 9.97 9.61 57.31 41.13 N 174 174 173 174 174 Including Hyperinflationary Countries Mean 18628.74 2.47 2.24 23.55 18.02 Median 18488.15 2.07 1.77 4.58 1.99 Standard Deviation 12636.68 1.55 1.63 123.08 114.77 Minimum 844.88 0.48 0.38 -0.08 0.31 Maximum 61974.36 9.97 9.61 1404.68 1216.68 N 182 182 180 182 182

Where: High Inflation is defined as countries with an average annual inflation rate of over 60%

(37)

33 high-income countries. This is not in line with data shown by Cecchetti et al (2005) and Calderón and Hebbel-Schmidt (2008). The primary reason for this is that the income categories for the countries included in the sample are not fixed. This implies that an individual country might have been included in different categories (depending on its GNI per capita) for different periods.

Table 5: Time Dynamics dependent variable (Excluding High Inflation Countries)

Period Descriptive Statistics Real GDP Per Capita GDP Growth Volatility % Consumption Growth Volatility % Inflation % Inflation Volatility % 1982-1988 Mean 16678.12 1.974 2.016 8.436 3.993 Median 20641.17 1.797 1.831 6.990 2.257 Standard Deviation 9889.717 0.948 1.070 6.838 4.575 Min 844.8782 0.846 0.498 1.476 0.917 Max 38811.68 5.163 5.107 37.361 21.514 N 30 30 30 30 30 1989-1995 Mean 18565.29 1.940 1.721 7.140 2.678 Median 21952.63 1.828 1.396 5.221 1.932 Standard Deviation 11404.51 0.871 1.208 5.586 2.590 Min 935.7532 0.581 0.380 1.736 0.565 Max 44880.84 4.583 6.710 24.675 12.290 N 34 34 33 34 34 1995-2002 Mean 18133.46 2.183 2.342 7.202 4.440 Median 14398.47 1.645 1.528 4.236 1.805 Standard Deviation 13022.15 1.419 2.028 9.956 7.524 Min 1020.087 0.477 0.395 -0.077 0.311 Max 55817.92 6.321 9.521 57.309 41.128 N 52 52 52 52 52 2003-2010 Mean 21425.73 3.07 2.32 4.41 2.07 Median 18396.46 2.54 2.03 2.99 1.55 Standard Deviation 14365.21 1.90 1.53 3.73 1.50 Min 1180.524 0.57 0.60 0.03 0.47 Max 61974.36 9.97 9.61 22.66 6.47 N 58 58 58 58 58

(38)

34 4.4. Robustness and Consistency of the Results.

As mentioned in section 4.1, I employ two proxies of volatile capital flows, one “total” measure and one measure that only incorporate foreign capital flows. In addition, I employ two different measures of export concentration, one that is pre calculated, taken from the IMF, and the Herfidahl index computed from UNCTAD trade statistics. I use these alternative measures to see if the results are consistent over different proxies for the same underlying variables.

(39)

35

V. Quantitative Results and Discussion

This chapter discussed the regressions results. Section 5.1 gives a general overview of the results and addresses how well the general assumptions of the regression models are met. Section 5.2 discusses the regression results for economic growth volatility, section 5.3 discusses the results for inflation volatility, and section 5.4 discusses the results for inflation volatility. Table 7 shows the main regression results. All regressions were significant with a probability of the F statistics smaller than 0.01%.

5.1. General Regression Results

In Appendix B, I included a table with the summary statistics for all regression data. The dependent variables are mostly skewed to the left. After normalising the variables by taking the natural logarithm, GDP growth volatility and consumption growth volatility are normally distributed. Although the shape of the distribution of inflation volatility improved significantly, it still passes a test for non-normality on a 99% significance level. Therefore the regression results for inflation volatility are less robust.

In Appendix C I included a correlation matrix. Capital account openness correlates heavily with credit as a percentage of GDP Real GDP per capita and Political rights. As expected, the two measures of export concentration (the primary) sourced from the IMF, and the secondary (Herfidahl index), have a correlation coefficient of 0.86. Furthermore inflation and inflation volatility also correlate heavily with a correlation coefficient of .86 (.75 excluding high inflationary countries).

(40)

36 exception of the regressions predicting inflation volatility where high inflation countries are included.

The fixed effects are significant for all regressions, which is as expected, as one cannot fully control for all country characteristics. 5.2. Economic Growth Volatility

Trade has, as hypothesised, no significant independent effect on economic volatility. When including its interaction effect with trade concentration, it only turns significant when including foreign capital flows in the regression (

Appendix D and D). However, the coefficient is consistently significant over both measures of trade concentration (Table 7). The interaction coefficient indicates that trade has a more negative effect for countries that have a higher degree of export concentration.

Table 6: Partial Elasticities for Trade Openness with respect export concentration

LOG EXPORT CONCENTRATION at: 10th Percentile 25th Percentile 50th Percentile 75th Percentile 90th Percentile ELASTICITES FOR TRADE % GDP -0.035 -0.186 -0.412 -0.607* -0.949** Theil index (IMF) (0.410) (0.370) (0.340) (0.343) (0.396) ELASTICITES FOR TRADE % GDP 0.273 0.061 -0.190 -0.632** -0.992*** Herfidahl index (0.410) (0.370) (0.340) (0.343) (0.396)

Where: The depended variable is economic growth volatility. Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1. Margins have been computed using the margins commando in Stata 13 using the regressions results of table 7: regression 2 and 4

As shown in the table above, the relationship between trade and economic growth volatility becomes only pronounced for higher values of export concentration. Trade openness has a pronounced stabilising effect for higher levels of export concentration. This is evidence against

H1.2 (The relationship between trade concentration and economic growth volatility is negative for countries with a high degree of export diversification and positive for countries with high concentrated exports). Therefore the corresponding null hypothesis could not be

(41)

37 Table 7: Main regression Results

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Standard deviation of growth in: GDP GDP GDP GDP CPI CPI CPI CONS. CONS. CONS.

LOG TRADE % GDP -0.446 0.411 -0.399 -2.428*** -0.127 -1.595** -1.022 0.0824 -0.312 -0.0413 (0.370) (0.515) (0.388) (0.804) (0.579) (0.783) (0.664) (0.311) (0.379) (0.394) LOG EXPORT CONCENTRATION 0.0107 5.339** 0.0487 2.768*** -0.802 -9.952*** -7.577*** 0.128 -2.410 -1.446 (0.507) (2.360) (0.134) (0.944) (0.960) (2.666) (2.374) (0.573) (1.638) (1.665)

LOG TRADE*LOG CONCENTRATION -1.275** -0.675*** 2.179*** 1.628*** 0.599 0.355

(0.599) (0.243) (0.653) (0.554) (0.423) (0.435)

LOG CAPITAL A/C VOLATILITY 0.180** 1.084*** 0.187** 1.230*** -0.229* -2.286*** -1.834*** -0.00339 -0.867** -0.757* (0.0859) (0.337) (0.0858) (0.372) (0.119) (0.442) (0.408) (0.0820) (0.381) (0.394) LOG CREDIT % GDP 0.241* 0.708*** 0.272* 0.846*** -0.217 -1.756*** -1.299*** -0.0439 -0.698** -0.644**

(0.141) (0.221) (0.145) (0.238) (0.200) (0.332) (0.324) (0.170) (0.306) (0.300) LOG CAPITAL A/C VOLATILITY*LOG CREDIT % GDP -0.227*** -0.265*** 0.559*** 0.432*** 0.235** 0.202**

(0.0850) (0.0967) (0.117) (0.109) (0.0964) (0.0990)

CAPITAL A/C OPENNESS -0.0306 -0.0509 -0.0365 -0.0231 -0.211** -0.181*** -0.0643 -0.0898 -0.0808 -0.0413 (0.0439) (0.0524) (0.0425) (0.0453) (0.0835) (0.0662) (0.0835) (0.0652) (0.0640) (0.0605) LOG RULE OF LAW -3.630*** -3.116*** -3.147*** -2.622*** -4.042*** -4.692*** -3.346** -4.201*** -4.404*** -3.709***

(0.996) (1.031) (0.982) (0.972) (1.292) (1.046) (1.352) (1.199) (1.144) (1.174)

POLITICAL RIGHTS 0.0359 0.0552 0.0540 0.0942 0.129 0.0741 0.0983 -0.0339 -0.0566 -0.0591

(0.0544) (0.0459) (0.0623) (0.0599) (0.0898) (0.0908) (0.0856) (0.0826) (0.0863) (0.0831)

LOG REAL GDP PER CAPITA 0.581* 0.659** 0.504 0.408 0.378 0.331 0.367 0.415 0.367 0.447

(0.335) (0.300) (0.339) (0.322) (0.460) (0.410) (0.386) (0.425) (0.431) (0.431) LOG INFLATION -0.0203 0.00762 -0.0143 0.0106 0.406* 0.123** (0.0509) (0.0536) (0.0485) (0.0504) (0.219) (0.0521) Observations 156 156 156 156 150 150 149 156 156 155 R-squared 0.197 0.243 0.189 0.251 0.369 0.424 0.490 0.175 0.204 0.232 Number of ID 54 54 54 54 54 54 54 54 54 54

Country Fixed Effects YES YES YES YES YES YES YES YES YES YES

Where: Red colored coefficients indicate that the secondary measure of a variable has been used. The primary measure for export concentration is the Theil

(42)

38 Capital account volatility is only significant when using the measure that

excludes domestic capital flows and including its interaction effect with the proxy of financial deepness in the regression (

Appendix D and D). When including the interaction effect, the coefficients of capital account volatility, financial deepness, and the interactionaction effect between the two become significant at a 99% significance level. The effects are consistent, also when excluding high inflation countries.

When including the interaction effect, the coefficients of capital account volatility and credit as a percentage of GDP are positive. The interaction effect is negative, which is consistent with hypotheses:

H1.3 Capital account volatility causes stronger economic growth volatility H1.4 The relationship between capital account volatility is more

pronounced for countries with weaker developed financial sectors and financial deepness.

When looking over the partial elasticities below, one notices that foreign volatile capital flows are in particular destabilising for countries with lower financial deepness. This is consistent with Caballero and Krishnamurthy (2001); Roderik 1999 and Aghion et al (1999).

Table 8: Partial Elasticities for Capital a/c volatility with respect to Credit as % of GPD

LOG CREDIT AS % OF GDP at: 10th Percentile 25th Percentile 50th Percentile 75th Percentile 90th Percentile CAPITAL A/C VOLATILITY 0.398*** 0.290*** 0.140* 0.052 -0.21

(0.106) (0.084) (0.082) (0.097) (0.115)

Where: The depended variable is economic growth volatility. Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1. Margins have been computed using the margins commando in Stata 13 using the regressions results of table 7: regression 2

Credit as a percentage of GDP has a consistent destabilising effect on economic growth volatility, including and excluding the interaction effect. A possible explanation could that credit as a percentage of GDP captures the effects of overleveraged financial sectors and credit bubbles that might precede financial crises.

(43)

39 literature. On the basis of this consistent insignificance, there is not enough evidence to reject the null hypothesis of:

H1.5 Capital Account openness reduces economic growth volatility.

Rule of law has a consistent stabilising effect on economic volatility with an average elasticity of 3% meaning that an percentage increase in the rule of law index on average implies a 3% increase in the standard deviation of economic growth.

5.3. Inflation Volatility

The regressions with inflation volatility as dependent variable differ significantly when excluding the eight outliers with high inflation. As a matter of robustness I therefore interpret the results excluding the outliers.

Trade as a percentage of GDP has, as hypothesised (H2.1), no independent effect on inflation volatility. When including the interaction effect of trade openness and export concentration, the three coefficients become significant. They are not significant over both measures of export concentration. When using the Herfidahl index of export concentration none of the effect is significant. In addition the effects become less significant and smaller when controlling for the average level of inflation (Appendix G).

The coefficients when including the interaction effect are consistent with the hypothesis. For a higher level of trade concentration the effect of trade becomes destabilising.

Table 9: Partial Elasticities for Trade % of GDP with respect to export concentration

LOG EXPORT CONCENTRATION at: 1st Percentile 10th Percentile 25th Percentile 50th Percentile 75th Percentile 90th Percentile 99th Percentile TRADE % OF GDP -1.241* -0.844 -0.567 -0.180 0.244 -0.784** -1.62*** Theil index (IMF) (0.692) (0.597) (0.538) (0.463) (0.408) (0.385) (0.476)

Where: The depended variable inflation volatility. Robust standard errors in parentheses *** p<0.01,

“On the relationship between openness and macro economic volatility: A cross sectional study”

Faculty of economics and business