The Effect of Financial Development on the Likelihood of Economic Crises to Occur

(1)

The Effect of Financial Development on the

Likelihood of Economic Crises to Occur

Master’s Thesis

Indra Figrachanda

S3117367

Supervisor : Dr. Pieter IJtsma

(2)

Abstract:

The focus of modern economics on finance is whether the development of finance would improve economic growth or cause crises. Studies found that there are indeed threshold of financial development until it becomes a deterrent of economic growth. With such a rich literature touching the subject there are surprisingly little empirical studies of the probability of economic crisis happening due to increasing financial development. This paper uses financial development variable to prove the hypothesis that financial development increases the probability and severity of economic crises. We find that financial development does have significant effect to the probability of economic crisis happening; however the effect for severity is only significant for a known financial crises period.

(3)

1. Introduction

Modern economics has a fickle relationship with finance; throughout history, finance has been a two-sided phenomenon when it comes to its role in affecting the economy. Scholars have been on either side of this debate, some have argued for financial development to its benefits in accelerating and increasing economic growth (Levine, 2003; King and Leving, 1993; Valickova, 2015). Others argue that uncontrolled increase of financial development is increasing the probability and the severity of economic crisis (Reinhart and Rogoff, 2009; Lane, 2012).

So does financial development helps promote growth or start crises? Studies have found that it can be both; financial development has been found to have a positive and negative impact on economic growth. Law & Singh (2014) and Arcand et al. (2015) worked on this theory and found that there is a threshold to the level of financial development until it becomes a hindrance instead of fuel to economic growth.

With the abundance of studies in this field, most that are focused on economic crises uses historical and comparative studies; there is a surprisingly little empirical study on the effect of financial development on economic crisis events. In this thesis, we would like to contribute to the field by proposing a set of empirical studies on this effect by posing two questions. First, do increases in financial development would increase the probability of economic crisis happening? And, does increases in financial development would increase the severity of economic crisis.

(6)

In the next section of this thesis (section 2), we will discuss with more depth theories about financial development, economic growth and crisis events that inspired this thesis as well as the hypothesis that we made. In section 3 we will elaborate our methodology to answer our two hypotheses using economic data that is widely available to the public. In section 4 we will discuss the results of our regressions as well some robustness check to ensure our chosen methodology was correct. Conclusions and some policy recommendations will be presented in the last section (section 5).

2. Theoretical Background

We will layout the theoretical background of this thesis as well as some related studies done by scholars before us in this section. We will discuss research and empirical studies that suggest financial development positively affect growth, and then studies concerning there is a threshold of financial development having positive effects on economic growth. Finally, we discuss the basis of our hypotheses and the theoretical background of it.

2.1. Finance: The Double-Edged Sword

Finance is considered an essential part of the economy in the modern era; evolution in the finance sector has created ample capital and credit that fueled the wheels of economies and spur growth. Scholars have made studies confirming this theory; King and Levine (1993) noted the strong relationship between financial depths, represented by Liquid Liabilities/GDP, to the pace of economic growth.

(7)

We have established that financial development plays a role in economic growth, as put forth by various analyses in the literature review part of this thesis. Valickova et al. (2015) found that financial development has played a significant role to promote economic growth since the well-developed financial market can provide more financing for development.

Furthermore, the microeconomic-based evidence is consistent with the view that better developed financial systems ease external financing constraints facing firms, which illuminates one mechanism through which financial development influences economic growth (Levine, 2005), Levine also suggests that financial development level can explain the differences in economic growth in countries (2003).

Another proxy for financial development is the financial innovation; a study by Laeven, Levine, and Michalopoulos (2015) finds that financial innovation is necessary to sustain economic growth. They analyzed the profit-maximizing effects of financial innovation to promote increased productivity, and seen positive effects of developments in financial markets to assist economic expansion.

A lot of these studies were done while looking at long-term time series of economic growth, but they fail to acknowledge the downturn period which may only be a blip in the series, but significant nonetheless as it could alter policies and regulations that will affect financial markets and economic growth.

As the ominous title of this section suggests, finance has also been noted to hurt growth. Thus it follows that financial development has a threshold in its impact on economic growth. Cecchetti et al. (2010): "Debt is a two-edged sword. Used wisely and in moderation, it clearly improves welfare. But when it used imprudently and in excess, the result can be a disaster…”. This statement suggests that there is a non-monotonic relationship between financial development and economic growth, a point echoed by Easterly, Islam, and Stiglitz (2000).

(8)

Rioja and Valev (2002) find that there are structural breaks on the effects of credit to economic growth. Testing several indicators, which includes private credits, and liquid liabilities they discovered that the effects of financial developments vary across different levels of a countries development status, and although overall it has positive effects, it will turn negative once the indicator gets too large.

These studies are consistent with Hyman Minsky’s Financial Instability Hypothesis which essentially describing that from the calmness of stability, there will rise innovators and with it financial innovations that rocks the stability from its safe equilibrium. Frederick Mishkin (2011) reinforce this theory by saying that with economic boom through increased consumption and investment and the rise in asset prices that causes over-confidence among investors is fanned by overflowing credit supply from financial institutions that forego prudential practices that finally lead to credit default and crises.

Price bubble is an economic event that can firmly explain the threshold theory. Theories on how the asset bubble is formed are closely related to credit and its evolution. Asset price growingly depended upon how much credit can be accessed instead of real demand and supply. The most recent and apt example is the housing market in the US circa 2000-2007. The expansion of money supply in a country that leads to the increased indebtedness of investors to purchase an asset that causes asset prices to soar above its real value (Gourinchas et al., 2001). As the further increase of debt will eventually cease, the popping of the bubble leaves investor's net worth to plummet and causes a crisis. This is the primary assumption that there is a level of financial development, which causes negative effects on economic growth.

This mechanism is what made the booms and busts in the economy where booms are indicated by exponentially increasing wealth and busts by the loss of value in the economy. Some busts are worse than others and may be associated with the fall of asset prices, such as noted by Kapp and Vega (2012) where debt crises in much more severe than currency crises. This asset price falls been increasingly monitored in the last two-decade where the tech-bubble of the 1990’s and of course the real estate bust in the early 2000s.

(9)

would follow suit flooding the shock with money. This new capital would fund the industries that started the shock and also the investors wanting a piece of the action, the investing frenzy will develop into a euphoria as the market goes up many times over and evolve into a mania until the capital flow stops and the market tanks (Loayza and Renciere., 2006; Aliber and Kindleberger, 2015).

2.2. Finance Effects of Economic Crises

Having already explained the effects of financial development on the economic growth and acknowledging that financial developments will have a threshold on its development to the positive growth level through the studies we have mentioned above, this thesis attempts to answer the research question of how financial development affects the probability of economic crises happening and the severity of them.

The base this thesis is that high debt level that is unsustainable which causes the collapse of economies over time. A factor in the model is the increasing credit supply over the course of the boom that fueled the balloon of debt and with it asset prices causing bubbles, and within it is ‘financial development.' Financial development has grown so fast during the past decade that it became too complex making risk assessment harder and contributed to causing economic crises (Demirguc-Kunt and Degatriache., 1998; Johnson and Kwak., 2012; McLeay at.al., 2014).

As Reinhart and Rogoff (2009) states: "When debt ratios rise beyond a certain level, financial crises become more likely and severe"; the first of our hypotheses: financial development increases the possibility of an economic crisis, is based on the model that exponential growth of credit-fueled asset price bubbles that lead to a crisis. We will use private Debt/GDP level as the dependent variable to prove this hypothesis as several other researchers that come before us (Arcand., 2012; Law and Singh., 2014, Loayza and Renciere., 2006; Chechetti and Kharroubi., 2012).

(10)

output losses that a country experience during economic crises and we will find how financial development affects them.

3. Methodology and Data

In this section, we will elaborate the method and data of our research that we introduced in Section 1. This section will be divided into two parts, Methodology, and Data. In the Methodology part, we will further explain our chosen definition for an economic crisis event, the models that we use to test our hypotheses, and some interpretation of previous studies that pertain to our endeavor. Also, we will elaborate our two hypotheses mentioned in earlier sections in this thesis. The Data part of this section we will define the data we chose as dependent, independent and control variables as well as the treatment that we apply to those data to do our analyses.

As we have mentioned in the first section of this thesis, surprisingly there is little empirical study that we come across during our research that looks into the relationship between financial developments to the probability of economic crisis happening. In this section, we give an alternative method to answer this question and use a logit model to predict probability as well as a different approach to determine an economic crisis event using data filter to create trends.

3.1. Methodology

3.1.1. Crisis Determination

Defining economic crises is crucial in our research, and there have been multiple approaches in the studies made by scholars before us. Most of the studies made on economic crisis have made use of economic surveys as well as historical annotation of a crisis event to define crisis periods (Summers, 2000; Hoggarth et.al., 2002; and Artha & De Haan, 2011), while we believe that this method is a definitive way to ensure an agreement on the period, but we opt to find a more empirical way to show an economic crisis happening.

(11)

movement of those variables to determine the economic crisis. While a lot of studies did not define the crises as a deep economic crisis, we take the understanding of their studies objects as economic crises. For example, to measure the depth and start of a currency crisis, Glick and Hutchison (2001) used the deviation of currency value against the trend from the moving average of said currency. Studies that measure banking crises mostly used real sectors variables such as GDP growth to proxy economic performance to identify the period of banking crises (Demirguc-Kunt and Degatriache, 1998; and Hutchison and McDill, 1999). Another measure for banking crises is to measure the degree to Financial Liberalization proxied by credit supply as done by Kaminsky and Reinhart (1996), they estimate the growth of credit supply over sample periods and define the start of crises when growth is below the trend established by moving average of the variable.

This thesis takes inspiration from those studies and tries to give a slightly different way to define an economic crisis. We define economic crisis as periods where the actual dependent variable, in this case, GDP per-capita falls below the GDP per-capita trend for two observation years in a row. We use economic data from the World Bank and cleaned the data by eliminating countries with lower than ten observations. Using panel data approach to this study, we apply a filter on the time-series data of our dependent variable to get the trend of the GDP per-capita variable. After filtering, we again eliminate countries with fewer than ten observations, and then we subtract the actual GDP per-capita data for each observation year. The result of the subtraction will determine whether the said year is in a boom or recession period and as we have explained, we define the periods where recessions happened at least twice in a row in our annual data as a crisis event, we thus designate this event with a dummy variable "1" and the non-event as "0".

(12)

3.1.2. Hypothesis I: An Increase in Financial Development Would Increase the Probability of an Economic Crisis Event Happening

Our first hypothesis, as was explained in previous sections, is that we believe that financial development significantly affects the probability of economic crises. We employ a logit model into two sets of equations attempting to prove this hypothesis, which we would further explain in the next three parts of this section. In the first part we will elaborate our method to test this hypothesis using a rather straightforward logit model into two equations, the first one assuming no time lag between the effects of Financial Development on the probability of an economic crisis event, and the second one with the assumption of a one year time lag on the effect. In the second part, we would investigate a further income effect affecting the original equations, where we will divide countries according to their income level and see if there are differences of effects affecting the results.

3.1.2.1. Financial Development Effect to the Probability of Economic Crises

After we find the crisis event for each sample country we then apply a logit equation determination where crisis period as we defined above are designated as dummy variable “1” and those that are not as “0”. This definition allows us to determine the effects of financial development on the likelihood of a crisis event happening. Using the logit model we were able to measure the probability of an economic crisis event occurring as a dependent variable to the financial development variable and several control variables.

(13)

Liability/GDP and Loan/Deposit Ratio (LDR) as we will explain in the Data part of this section.

Model (1) shown below is the model that we use to prove our first hypothesis, which is; “financial development increases the probability of an economic crisis event happening."

𝑃𝐶!" = 𝛽!+ 𝛽!𝐹𝐷!" + 𝛽!𝜒!" + 𝜀 (1)

Where PC = _{0 𝑖𝑓 𝑒𝑐𝑜𝑛𝑜𝑚𝑖𝑐 𝑐𝑟𝑖𝑠𝑖𝑠 𝑒𝑣𝑒𝑛𝑡 𝑑𝑖𝑑 𝑛𝑜𝑡 𝑜𝑐𝑐𝑢𝑟}1 𝑖𝑓 𝑒𝑐𝑜𝑛𝑜𝑚𝑖𝑐 𝑐𝑟𝑖𝑠𝑖𝑠 𝑒𝑣𝑒𝑛𝑡 𝑑𝑖𝑑 𝑜𝑐𝑐𝑢𝑟

Where 𝑃𝐶_!" is a dummy variable explained above where “1” is when an economic event occurs, and “0” is when the economic crisis event doesn't occur. True to our hypothesis, the probability of economic crisis event depended upon the level of financial development, which is represented by 𝐹𝐷_!" in our model. If our hypothesis is proven, then we expect to see a positive and significant relationship between financial development and the probability of economic crisis event happening, which means that continued growth in financial development would increase the probability of an economic crisis occurring. At this point, we haven't talked about measures of financial development that causes the severity of the economic crisis we just wanted to see how the development of financial sector affects the fragility of an economy to the point that it increases the possibility of an economic crisis. We included some control variables to this equation to add conduciveness to the result, and 𝜀 denotes an error term.

In testing this hypothesis, we also check another suspicion that the effects of financial development lag on the probability of economic crisis event happening. This analysis is due to our inference that financial development that affects economic crises is the financial development that occurred before the economic crises began. Using this definition we lagged the data used to regress the model by one period from the crisis event and see the effects of prior financial development to the crises that eventually happened and does said development causes the economic crises that ensue. The equation used is thus termed in this thesis as Model (2).

(14)

As before 𝑃𝐶_!" denotes the same economic crises definition as Model (1) while 𝐹𝐷_!!! denotes the financial development level that is one year before the economic crises event, and 𝜀 is still denoting the error term. As before if our hypothesis is correct, we expect a positive and significant relationship between financial development and the probability of an economic crisis event are happening.

3.1.2.2. Income Effect to the Probability of Economic Crisis Happening

In this section, we aim to look at the difference between the financial development effect on economic crises between developing and developed markets. This analysis is inspired by the work of Rioja and Valev (2002) that found differences in the impact of financial developments on the economic growth of different classification of economic development countries.

With this base, we aim to see if there are differences of effects that financial development brings to economic crises in a different classification. Rioja and Valev used the World Bank classification of income to make their case and used the designation for high, medium and lower income countries. We also used the World Bank classification, but since we are aiming to analyze the effects of financial development on high-income nations compared to the rest-of-the-world, we are only going to take the High-Income countries from the data as the sample for High Income and used the rest of the data as Rest-of-the-world. The reasoning behind this is that global crises during 1990-2016 have occurred in financially well-developed countries. We are curious whether the financial development plays a significant role in the crises event of that country.

(15)

To do this analysis, we introduced another dummy variable representing the two types of nations that we divide-up based on its income level. We used dummy variable “1” to denote high-income countries and “0” to denote non-high-income countries. We include this dummy variable in our model with its interaction with financial development variable as noted in Model (3).

𝑃𝐶!" = 𝛽!+ 𝛽!𝐹𝐷!" + 𝛽!𝐼 𝐹𝐷!"+ 𝛽!𝜒!" + 𝜀 (3)

Where Ι = _{0 𝑖𝑓 𝑛𝑜𝑛 − ℎ𝑖𝑔ℎ − 𝑖𝑛𝑐𝑜𝑚𝑒 𝑐𝑜𝑢𝑛𝑡𝑟𝑖𝑒𝑠}1 𝑖𝑓 ℎ𝑖𝑔ℎ − 𝑖𝑛𝑐𝑜𝑚𝑒 𝑐𝑜𝑢𝑛𝑡𝑟𝑖𝑒𝑠

If our hypothesis about income effect holds, then we should find a significant coefficient for the interaction term between I and 𝐹𝐷!". Should we find the expected

result, we then look at each country group separately and apply Models (1) and (2), to see the differences of the 𝐹𝐷_!" coefficient in high-income and non-high-income.

3.1.3. Hypothesis II: An Increase of Financial Development Increases the Severity of Economic Crisis.

In our effort to prove our second hypothesis; “An increase of the degree of financial development would affect the severity of the economic crisis event," we look to several studies that quantify output loss in the event of a crisis. Some studies calculated the output loss by subtracting the peak and trough of the GDP per-capita during a known economic crisis event (Artha and De Haan, 2011; Kapp and Vega, 2012). The economic crisis event determination varied across studies with some using quarterly GDP per-capita data to note a consecutive negative growth as an economic crisis, and others used expert opinions and surveys to determine an economic crisis period.

(16)

To prove our hypothesis we use a simple panel data regression using Output Gap, measure of how much GDP per-capita deviate from its trend, as the dependent variable. We use positive and negative output gaps in this analysis to eliminate selective data bias that would occur if we just include the negative ones. To prove our hypothesis we apply Model (4) shown below where output gap is used as dependent variable and if our hypothesis is proven we should find a negative and significant result in our regression.

𝑂𝐺_!" = 𝛽_!+ 𝛽_!𝐹𝐷_!" + 𝛽_!𝜒_!" + 𝜀 (4)

We also used the logic that produced Model (2) from the probability analysis by lagging the independent variable for one period. This Model is shown below and designated Model (5).

𝑂𝐺!" = 𝛽!+ 𝛽!𝐹𝐷!"!!+ 𝛽!𝜒!"!! + 𝜀 (5)

The dependent variable of this denoted by 𝑂𝐺_!" is the output gaps, which is derived by subtracting the GDP per-capita number from the trend. As in previous equations, 𝐹𝐷_!" is the financial development variable, and 𝜒_!" denotes the control variable. To prove our hypothesis we expect to see a positive and significant coefficient for 𝐹𝐷_!" variables as an effect on the output loss in an economic crisis.

3.2. Data

3.2.1. Dependent Variable for Probability of Economic Crisis Event

(17)

countries due to incomplete data; this leaves us with 180 countries with 6,484 observations.

To get the dummy variable, we applied several filtering to our dataset to acquire the GDP per-capita trend that is crucial for its formation. We used two filtering method, the Hodrick-Prescott and Baxter-King filters, to acquire the trend. The validity of this technique is shown in Appendix 1 where we apply the filter to the data of US GDP per-capita, United Kingdom, Germany, and Japan. We used the full dataset from 1961 to 2016 as well as only until 2006 to see whether the filter worked to identify crisis event as noted by each country. We can see that the filter correctly identify several noted economic crisis events in both datasets and from all countries. This result empowered our method of ultimately defining an economic crisis event that is used throughout this thesis.

3.2.2. Dependent Variable for Output Loss in an Economic Crisis Event

For this analysis, as we explained in the previous part, we used the nominal output loss from each of the economic crisis events as provided from the study we did earlier to define an economic crisis. As a result, the GDP per-capita nominal that we take for this analysis is only those that is less than zero after we subtracted them from the trend. This method would adequately resolve our output loss definition and would fit the model to prove the second hypothesis of this thesis.

3.2.3. Dependent Variable to Test Income Effect

We used the same dataset for the income effect testing and then separate the country groups based on the definition of high-income countries by OECD, for this analysis, we only separated the high-income countries from the non-high-income countries. We found 52 countries that are categorized as high-income by OECD with 2,006 observations while the rest-of-the-world group has 131 countries with 4,588 observations.

(18)

3.2.4. Independent Variable

The independent variable used in this thesis is the degree of financial development with which we want to understand its effect to the economic crises. As a proxy for financial development, we use the most common indicator in most studies concerning financial development to economic growth and economic crises, which is the credit to the private sector as a percentage to the GDP.

Admittedly this thesis has tried several other variables to proxy financial development. We try to use the LDR (Loan to Deposit Ratio) which we take as one of the definitions of financial development since a high LDR would indicate that loan in the economy is not funded by deposit instead of by the market instrument which is the base of market-based banking. Due to availability issues in the dataset, we would loose as much as half of our observation and thus we decided against using this variable until we can obtain a more complete data.

Another variable we tried compiling is the Liquid Liability/GDP, which is another variable used to define financial development is many studies (Arcand, 2012., Law and Singh., 2014., Levine (2008), is the sum of liquid money that is in circulation in an economy. However, we find incomplete data from all of our samples and some calculations have to be manually made in order to have a more complete data, even then we loose about 14 countries and 500 observations. Due to worry about inaccuracies in our calculations to represent the real conditions of an observation we decided against using this variable as our main explanatory variable, however we do regressed our Models using this variable as a robustness check at the end of this thesis.

Since private credit data is readily available and currently the most reliable proxy for financial development, we decided to use this variable to prove the hypothesis. We collected the data from World Bank database of economic indicators.

3.2.5. Control Variables

(19)

We used Trade as a percentage to GDP as a proxy for economic openness under the thesis that an open economy would increase foreign currency transactions and money supply that affect economic growth and probability of a crisis. Another control that we used in Population Growth under the thesis that an increased population affected economic growth. The last control variable we used is the inflation; we use this variable under the thesis that inflation affected money growth in an economy and hence affected economic expansion and recession.

All of this data was obtained from OECD annual data and were combined as a data set to the dependent and independent variables. As the dependent and independent variables, we found several missing data points, but after we give the treatment as stated in the first section of the methodology data, we have a somewhat balanced sample.

4. Results and Analysis

In this chapter, we will discuss our findings after we did our regression on the model defined in Chapter 3. We find the result somewhat consistent with our hypothesis that there is a significant effect of financial development to economic crises. This chapter is divided into several sections to answer several questions in our methodology chapter.

4.1. Financial Development Effect on Economic Crises

In this section, we will discuss more of Model (1) where the financial development variable used as independent and economic crises events as the dependent variable is in the same period. Table 4.1 columns 1 and 2 are the regression results for Model (1).

(20)

Although this is consistent with our hypothesis that financial development has a significant effect on the economic crises event observed, we calculated the probability of an economic crisis event happening using coefficients from column 2 and found that at 85%, that is one of the critical levels discussed by Law & Singh (2015), the probability of an economic crisis happening still stands at 39.6% and every 5% increase of Private Debt-to-GDP will result in a 0.35% increase of probability.

We also found that one of our control variables, which is Trade-to-GDP has a negative and significant effect on a 99% confidence level to the possibility of a crisis event happening. We took this to mean that contrary to some studies that find the more open an economy, the more susceptible they are to a crisis event, the economic crises would be more probable to happen to a closed economy rather than an opened one. The probability of crises happening would decrease as the country increases its trade, although true to the theory heterogeneous effect of finance due to trade imbalance (Barro., 2001; Cetorelli and Goldberg., 2011; Borio and Disyatat., 2010), an increasing trade effect of decreasing crises probability lessen for every 5% increase in trade.

4.2. Financial Development as a Lagging Indicator

As for the model that suggested finance development indicator as a lagging indicator to the logit model of an economic crisis event (Model 2), we find a positive and significant effect of the financial development effect to the possibility of economic crisis event happening. We analyzed this using Model (2), and after Hausman testing, we still find that using fixed effect is more efficient in estimating this model, thus as discussed in the previous chapter as shown in Table 4.2. Below in columns 1 and 2, and as we did before we apply the fixed effects in section 2.

(21)

than the current one. We took this to mean that activities from the past indeed affected the current outcome greater than current activities.

Table 4.1.

Estimation of Model (1)

This table shows the logit regression result from Model (1) where we imply that crises potentiality is affected by private debt/GDP ratio. We added some control variables that may affect crises potentiality as referenced by other authors such as; trade/GDP, population growth, and inflation.

(1) (2)

VARIABLES Crises Crises

Private Debt/GDP (%) 0.00153** 0.00292** (0.000748) (0.00131) Trade/GDP (%) -0.000969* -0.00670*** (0.000547) (0.00145) Population Growth -0.0413* -0.000956 (0.0218) (0.0417) Inflation 0.000133* 0.000160* (7.98e-05) (9.09e-05) Constant -0.684*** (0.0698) Observations 6,484 6,484 Number of countries 180 180

Fixed Effects YES

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

After calculating Probability of crises happening from the fixed effects model, we found that similar to Model (1), Private Debt-to-GDP level at 85% creates a 40% probability of an economic crises event happening. Also similar to Model (1) we found that an increase of 5% in Private Debt-to-GDP level will increase the probability of an economic crisis event happening by 0.5%, this is then consistent with the authors we cite in previous section of this thesis that higher indebtedness does make financial crisis more likely (Reinhart and Rogoff., 2009).

(22)

negative and significant result in the logit regression. A probability calculation found that its effect is slightly higher in this Model than in Model (1) with a 5% increase of Trade-to-GDP levels decrease crisis potentiality by 0.5% with a decreasing marginal effect.

Table 4.2.

Estimation of Model (2)

This table shows the logit regression results from Model (2) of our thesis where we posit that crises probability of a given year is depended upon several variables having been lagged for a year. The independent variables of this model are the same as Model (1), but all have been lagged for one year.

(1) (2)

Private Debt/GDP (%) 0.00237*** 0.00458*** (0.000753) (0.00131) Trade/GDP (%) -0.000938* -0.00606*** (0.000545) (0.00144) Population Growth -0.0115 0.0957** (0.0216) (0.0420) Inflation 0.000117 0.000139* (7.27e-05) (8.10e-05) Constant -0.770*** (0.0695) Observations 6,484 6,484 Number of countries 180 180

Fixed Effects YES

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

4.3. Financial Development in High-Income Countries

(23)

this thesis we then interact the dummy variable with the Financial Development Variable and see whether the coefficient is significant for us to determine there are income-effect in the model. The result of the logit regression is shown in Tables 4.3.1 and 4.3.2. below.

Our result for the Model (1) regression with the identifier shows that there is no income-level effect for the Model as shown by the result in the third row of column 1 with the financial development variables retaining its significance albeit at 90% confidence level. We then do another regression only with the financial development model and its interaction with the income level identifier to see if its significance level change should the control variables are left out, the results show that the interaction variable has no significance level even though its magnitude increases. Hence, we conclude that in our probability analysis of economic crisis event happening, income level does not play a significant part in affecting it.

During our analysis in Section 1, we found that Trade/GDP variable has great significance in affecting the probability of economic crisis event happening. Following the logic that countries with high-income level have higher trade values, we postulate that there will be an income effect to Trade/GDP as a result shows in column 1 the interaction effect is not significant. We also regress the model with only Trade/GDP, Income Level Identifier, and their interaction as we did for Private Debt/GDP and the result still holds. Therefore we conclude there is no income effect in Model (1) of this thesis.

We did the same analysis for our Model (2), the result of which is on Table 4.4., we found the same result for this analysis with Interaction Variables of Income Level Identifier with Private Debt/GDP and Trade/GDP holds no significance to the model, leading us to the conclusion that there is no income-effect relating to the possibility of economic crisis event in Model (2) the same as our finding in Model (1).

What we find interesting was that in Model (2) our main variable, Private Debt/GDP, gives better significance level than in Model (1) with Income Level variables added. This result reinforces our belief that predicting a probability of economic crises happening will be better if all explanatory variables are lagged.

(24)

robustness check. We show our results in Appendix 2 of this thesis, and we analyze them in the ‘Robustness Check' part of this section.

We also regressed Models (1) (2) again, with split samples created by splitting the countries along the high-income line set by OECD for our robustness check. We show our results in Appendix 2 of this thesis, and we analyze them in the ‘Robustness Check' part of this section.

4.4. Effects of Financial Development on the Severity of the Crises

The second hypothesis in this paper is an increase in financial development, will increase the output loss in times of economic crises. As we stated in Section 3 of this paper, we used a Panel Data estimation model to prove this hypothesis, the result of which is shown in Table 4.4.1. As an indication of economic crisis severity, we calculated output gaps from the deviation of GDP per-capita at a given time to the trend or potential GDP per-capita of the same year. Using this as dependent variable to the financial development independent variable along with controls we aim to prove our second hypothesis.

The result fails to find significance in the effect of financial development to the severity of the crisis. In fact in Model (4) results we do not find significance on the financial development variable to the output gaps, however, after we focused our research to the year 2006-2016, a period known to include financial crisis events, we fine negative and significant effect of financial development to output gaps. The result shown in the first row of column 2 is interpreted that a 1% increase of financial development will negatively affect output gaps by 0.01% across all country samples. Because the result for financial development effect was negative and significant to output gaps, we then conclude that increase in financial development in the period from 2006 onwards, increases output loss during the period and thus increases the severity of an output loss.

Table 4.3.1

Interaction of Income Level Dummy with Model (1)

(25)

along with the identifier and the interaction between the identifier dummy with Trade/GDP.

(1) (2) (3)

VARIABLES Crises Crises Crises

(26)

Table 4.3.2

Interaction of Income Level Dummy with Model (2)

Almost identical for what we did in Table 4.3, in this table we show the results for our regression of Model (2) adding the income level identifier dummy along with its interaction with the lagged Private Debt/GDP and Trade/GDP variables.

(1) (2) (3)

VARIABLES Crises Crises Crises

Private Debt/GDP (%) T-1 0.00711*** 0.00378* (0.00218) (0.00209) Income Level Identifier (Dummy) - - - Interaction Variable (Private Debt/GDP T-1 * Income Level) -0.00370 -0.00235 (0.00259) (0.00257) Trade/GDP (%) T-1 -0.00620*** -0.00551*** (0.00145) (0.00182) Population Growth T-1 0.1000** (0.0422) Inflation T-1 0.000141* (8.13e-05) Interaction Variable (Trade/GDP T-1 * Income Level) 0.00184 (0.00290) 0.00168 (0.00266) Observations 6,594 6,594 6,594 Number of countries 183 183 183 Fixed Effects Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.

While rerunning the Model to period before 2006, we found Trade to be the more significant variable among the independent variables. We conclude that financial development would only have significant effect to output gaps in the financially turbulent period, where the world is reeling from the effect of financially led shocks which is what the period 2006 onwards is defined by.

(27)

financial development is not statistically significant, having only P-value of 0.0012. However just as we did with Model (4) we reran the Model using data for the year 2006 onwards and data from before the year of 2006. We found that financial development is statistically significant on both periods as shown in columns 2 and 3 respectively. The financial development effect is economically more significant for the period for the years 2006 onwards than for periods before 2006. For the years of 2006 onwards an increase of 1% in financial development will negatively impact the output gap by 0.02% while for the periods before 2006, the effect is significantly lower at 0.006%.

With these results, we stand by our initial reading of this hypothesis test that financial development will only significantly affect output gaps, and therefore potential economic crises, in times of known crisis periods; consistent with the view that increased indebtedness affects severity of crises (Reinhart and Rogoff., 2009).

Thus our findings are inconclusive to prove the second hypothesis of this paper. Generally, we find no significant effect of financial development to the severity of crises, but we do find a significant result in the periods from 2006 onwards.

4.5. Robustness Check

In this section we check the robustness of all of our models, we divide this section into two parts. We check the robustness of Models (1) and (2) by splitting the country samples and by running the models with data from before the global crisis of 2006-2007.

4.5.1. Difference of Financial Development Effect based on Income Level of Country Samples

(28)

observations and 131 countries for the non-high-income group with 4,588 observations, so we believe both sets of data is quiet enough to represent the model.

Table 4.4.1.

Financial Development Effect on Output Gaps

This table shows the result from Model (4) estimation. Column 1 is the result using data from all time periods of the data. Column 2 is the result of using data from the years 2006-2016, where there is known financial crisis events were happening. Column 3 is the result of using the data from the period before the year 2006.

(1) (2) (3) VARIABLES Output Gaps All Period Output Gaps >=2006 Output Gaps <2006

Private Debt/GDP (%) 7.62e-05 -0.0161*** 2.89e-05

(0.00135) (0.00611) (0.00253) Trade/GDP (%) -0.000498 0.0106** -0.0104*** (0.00112) (0.00540) (0.00254) Population Growth -0.0699* -0.0315 -0.182*** (0.0382) (0.126) (0.0675) Inflation -0.000379*** 0.0388*** -0.000362*** (8.72e-05) (0.0143) (9.15e-05) Constant 0.134 -0.0123 0.940*** (0.136) (0.648) (0.228) Observations 6,484 1,702 4,782 Number of countries 180 178 175 Table 4.4.2.

Financial Development Lagging Effect on Output Gaps

(29)

(1) (2) (3) VARIABLES Output Gaps All Period Output Gaps >=2006 Output Gaps <2006 Private Debt/GDP (%) T-1 -0.00207 -0.0265*** -0.00626** (0.00137) (0.00606) (0.00262) Trade/GDP (%) T-1 0.00242** 0.00613 0.00271 (0.00113) (0.00516) (0.00258) Population Growth T-1 -0.120*** -0.135 -0.264*** (0.0382) (0.129) (0.0699) Inflation T-1 -0.000258*** -0.0752*** -0.000227** (8.48e-05) (0.0146) (8.93e-05) Constant 0.0693 1.736*** 0.401* (0.136) (0.607) (0.233) Observations 6,484 1,702 4,782 Number of countries 180 178 175 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

We found that financial development variable still retains significance for both types of samples with almost the same way with when we regressed them using all of the samples available. Table A2.1. In Appendix 2 show our regression Model (1) for both types of samples. We found the same level of significance for both samples sets with non-high-income countries having higher magnitudes. After we calculate for probability, we found that at 85% Private Debt/GDP level the initial probability of economic crisis happening is almost identical for both sets of data at 39%. However, after an increase of 5% in the variable, we found that the probability of economic crises happening increased higher in non-high-income countries than the high-income countries, with non-income countries increasing probability by 0.48% and high-income countries increasing by 0.32%.

(30)

higher for non-high-income countries; at 85% Private Debt/GDP level, we found that for non-high-income countries the probability of an economic crisis happening is 46.7% versus that of high-income countries which are 38.6%. Furthermore, the increased probability of a 5% increase in Private Debt/GDP for non-high-income countries is close to 1% versus that of high-income countries of 0.41%.

Using the logic that high-income countries are likely to have a better developed financial system than non-high-income countries, we think that this result is due to lack of infrastructure, regulation in non-high-income countries, or just that the system in non-high-income countries are less developed than those of its counterpart.

The results of this regression show that our result is still robust for different types of samples.

4.5.2. Financial Development Effect from Period Before the Great Financial Crisis of 2006-2007

The great financial depression of 2007 was the epitome of financial development affecting an economic crisis event. Financial development regarding financial innovation in debt markets and unprecedented private debt levels makes economies more fragile. The world, particularly high-income countries, reeled from the US Sub-prime mortgage crisis that causes an economic depression. Not having fully recovered from the depression, the Eurozone sovereign debt crisis happens in 2011 putting periphery Eurozone countries into deep recession. To check if our model works in predicting probability in the period before financial development started an economic crisis, we test our models for the period before the economic crisis by omitting every sample from the year 2006 onwards.

(31)

Private Debt/GDP level the probability increases by an average of 1%, this is also much higher than the initial results from this Model.

The lagging model results are shown in Table A3.2. Tell a more or less the same story with increased significance and magnitude for the financial development variable. The probability calculation shows an even higher probability of economic crisis at 85% Private Debt/GDP level with 56.8%, and a 5% increase of Private Debt/GDP increases the probability by 1.4%.

This result shows how the increase Private Debt/GDP during this era has sparked warning signs of economic crisis signaling that an economic crisis is more than 50% likely to happen. We have a suspicion that if we regressed the model with data from countries that were affected the recession of 2007 using the same period, we would find an increased probability of economic crisis happening. However, due to lack of time to instigate another method of choosing the samples, we would not regress that in this thesis but would like to do a more in-depth research on that in the future.

The results of this robustness check show that our initial findings are still robust for the different period.

4.5.3. Financial Development Effect using Different Explanatory Variable

As we discussed in Section 3 of this thesis, we used one more explanatory variable to proxy financial development, which is Liquid Liability/GDP. As shown in table A4.1 in Appendix 4, we still found significant effect of financial development in increasing the probability of economic crises happening using Model (1) and Model (2).

(32)

What is different between using this variable than Private Debt/GDP, is that financial development effect is clearly the only one significant statistically and economically from all the explanatory variables.

5. Conclusions

We offered two hypotheses in this paper, the first of which is that an increase of financial development will increase the likelihood of economic crises; the second one is that an increase in financial development will increase the severity of economic crises.

Using logit model we have proven our first hypothesis that financial development positively and significantly affect the probability of economic crisis events. We found that increased financial development, using Private Debt/GDP as a proxy, would increase the likelihood of economic crisis event happening. This result is robust for high-income countries and non-high income countries with no significant difference between the two income groups.

Our findings fail to conclusively prove our second hypothesis after we only found financial development only significantly affect output gaps during economic crisis periods, which in this paper we define as the period of 2006-2016. Although we do find a statistically significant result for financial development effect to output gaps for periods before 2006, it is economically non-significant therefore we still cannot conclusively prove our second hypothesis.

Another finding of our research is that financial development is a lagging indicator rather than current. Across our findings, it is increasingly clear that financial development for the period before an economic crisis event has a significant effect on the eventual crisis happening. This finding is robust for different income groups as well as periods.

(33)

Using the results of our regressions, we propose some policy recommendation to evade another devastating financial crisis. First, it is clear that financial development has an important effect on economic growth and exponential increase in financial development would increase the probability of a crisis. Therefore government should increase regulations on private credit to slow private credit growth so as not to increase faster than economic growth. Second, we propose to slow the growth of the financial institution; this move is yet another effort to reduce the pace of private credit growth in an economy. Finally, responding to the result that financial development negatively affects output gaps in times of crises, we propose that during financial crisis government should break-down big financial institutions to slow credit growth to reduce its lag to economic growth.

The limitation of this thesis is the data availability for other variables that we think could proxy for financial development. There are far too many variables left unaccounted for in this stylized facts to prove hypotheses, much of it has to do with the continually evolving world of finance, and more of that evolution are not yet recorded in data or simply not shown in the publicized one.

In our opinion, there are several things that we would have liked to do with our current study. We would have like to see how financial innovation affected the financial development in different income-level countries and we also would have liked to have more variables to proxy financial development to get a wider if not better picture for the effect that financial development has in causing financial crises.

We believe that the contribution of this thesis gives one more empirical evidence toward the effect of financial development has toward economic crises. There is an application of this study for policy-making purposes, but we hope to do a more in-depth review of financial development in its effects toward an economic crisis. Further development of this research would possibly lead to the development of an index of financial development variables in for a complete early warning system for economic crises, which we aim to develop further in the future.

(34)

References:

1. Arcand, J. L., Berkes, E., & Panizza, U. (2015). Too much finance?. Journal of Economic Growth, 20(2), 105-148.

2. Artha, I. K. D. S., & de Haan, J. (2011). Labor market flexibility and the impact of the financial crisis. Kyklos, 64(2), 213-230.

3. Barro, R. J. (2001). Economic growth in East Asia before and after the

financial crisis (No. w8330). National Bureau of Economic Research.

4. Bezemer, D.J. (2011). Debt, the Good, the Bad and the Ugly. University of Groningen. https://www.youtube.com/watch?v=nyGKs-M8SeU

5. Borio, C., & Disyatat, P. (2010). Global imbalances and the financial crisis: Reassessing the role of international finance. Asian Economic Policy

Review, 5(2), 198-216.

6. Cecchetti, S. en GE Kharroubi, 2012, Reassessing the impact of finance on

growth. BIS Working Paper 381.

7. Cecchetti, S., Mohanty, M., & Zampolli, F. (2011). Achieving growth amid fiscal imbalances: the real effects of debt. In Economic Symposium

Conference Proceedings (Vol. 352, pp. 145-96).

8. Cetorelli, N., & Goldberg, L. S. (2011). Global banks and international shock transmission: Evidence from the crisis. IMF Economic Review, 59(1), 41-76.

9. Dekle, M. R., & Pradhan, M. (1997). Financial liberalization and money

demand in ASEAN countries: implications for monetary policy. International

Monetary Fund.

10. Demirguc-Kunt, Asli., & Degatriache, G. (1998). The Determinants of Banking Crises in Developing and Developed Countries. International

Monetary Fund Staff Papers 45, 81-109.

11. Demetriades, P. O., & Hussein, K. A. (1996). Does financial development cause economic growth? Time-series evidence from 16 countries. Journal of

development Economics, 51(2), 387-411.

12. Easterly, W., Islam, R., & Stiglitz, J. E. (2001). Shaken and stirred: explaining growth volatility. In Annual World Bank conference on development

(35)

13. Friedman, M., & Schwartz, A.J. (1963). A Monetary History of the United

States: 1867-1960. Princeton University Press.

14. Frydl, M. E. J. (1999). The length and cost of banking crises. International Monetary Fund.

15. Girgin, S. C., Nguyen, H. O., & Karlis, T. (2017). Revisiting the Effect of Financial Development on Economic Growth after the 2008 Global Financial Crisis.

16. Glick, R., Moreno, R., & Spiegel, M. M. (Eds.). (2001). Financial crises in

emerging markets (No. 2001-2007). Cambridge University Press.

17. Gourinchas, Pierre-Olivier, Rodrigo Valdes, and Oscar Landerretche. Lending

booms: Latin America and the world. No. w8249. National Bureau of

Economic Research, 2001.

18. Greenwood, R., & Scharfstein, D. (2013). The growth of finance. The Journal

of Economic Perspectives, 27(2), 3-28.

19. Gupta, R. (2007). Financial liberalization and inflationary dynamics: an open economy analysis. International Economic Journal, 21(3), 335-360.

20. Hoggarth, G., Reis, R., & Saporta, V. (2002). Costs of banking system instability: some empirical evidence. Journal of Banking & Finance, 26(5), 825-855.

21. Hutchison, M., & McDill, K. (1999). Are all banking crises alike? The Japanese experience in international comparison. Journal of the Japanese and

International Economies, 13(3), 155-180.

22. Innes, A. M. (2004). The credit theory of money. Credit and State Theories of

Money: The Contributions of A. Mitchell Innes, 50.

23. Johnson, S., & Kwak, J. (2012). Is Financial Innovation Good for the Economy?. Innovation Policy and the Economy, 12(1), 1-16.

24. Kaminsky, G., Lizondo, S., & Reinhart, C. M. (1998). Leading indicators of currency crises. Staff Papers, 45(1), 1-48.

25. Kaminsky, G. L., & Reinhart, C. M. (1999). The twin crises: the causes of banking and balance-of-payments problems. American economic review, 473-500.

(36)

27. Kindleberger, C. P., & Aliber, R. Z. (2001). Anatomy of a Typical Crisis. In Manias, Panics and Crashes (pp. 21-32). Palgrave Macmillan UK.

28. King, R. G., & Levine, R. (1993). Finance and growth: Schumpeter might be right. The quarterly journal of economics, 108(3), 717-737.

29. Laeven, L., Levine, R., & Michalopoulos, S. (2015). Financial innovation and endogenous growth. Journal of Financial Intermediation, 24(1), 1-24.

30. Laeven, L., & Valencia, F. (2013). Systemic banking crises database. IMF

Economic Review, 61(2), 225-270.

31. Lane, P. R. (2012). The European sovereign debt crisis. The Journal of

Economic Perspectives, 26(3), 49-67.

32. Law, S. H., & Singh, N. (2014). Does too much finance harm economic growth?. Journal of Banking & Finance, 41, 36-44.

33. Levine, R. (2003). More on finance and growth: more finance, more growth?. Review-Federal Reserve Bank of Saint Louis, 85(4), 31-46.

34. Levine, R. (2005). Finance and growth: theory and evidence. Handbook of

economic growth, 1, 865-934.

35. Loayza, N. V., & Ranciere, R. (2006). Financial development, financial fragility, and growth. Journal of Money, Credit and Banking, 1051-1076. 36. McLeay, Michael, Amar Radia, and Ryland Thomas. "Money creation in the

modern economy." (2014).

37. Mehrling, P. (2013). 21 The inherent hierarchy of money. Social Fairness and

Economics: economic essays in the spirit of Duncan Foley, 169, 394.

38. Minsky, H. P. (1992). The financial instability hypothesis.

39. Mishkin, F. S. (1992). Anatomy of a financial crisis. Journal of evolutionary

Economics, 2(2), 115-130.

40. Mishkin, F. S. (1999). Lessons from the Asian crisis. Journal of International

Money and Finance, 18(4), 709-723.

41. Mishkin, F. S. (2011). Monetary policy strategy: lessons from the crisis (No. w16755). National Bureau of Economic Research.

42. Papadimitriou, D., & Wray, L. R. (2010). The Elgar Companion to Hyman

Minsky. Edward Elgar Publishing.

(37)

44. Reinhart, C. M., & Rogoff, K. S. (2009). This time is different: Eight centuries

of financial folly. princeton university press.

45. Reinhart, C. M., & Rogoff, K. S. (2009). The aftermath of financial crises. American Economic Review, 99(2), 466-72.

46. Rioja, F., & Valev, N. (2004). Does one size fit all?: a reexamination of the finance and growth relationship. Journal of Development economics, 74(2), 429-447.

47. Schularick, M., & Taylor, A.M. (2010). Credit Booms Gone Bust: Monetary

Policy, Leverage Cycles and Financial Crises, 1870-2008. Leibniz

Information Center for Economics.

48. Schumpeter, J. (1928). The instability of capitalism. The economic

journal, 38(151), 361-386.

49. Shrestha, M. B., & Chowdhury, K. (2006). Financial liberalization index for Nepal.

50. Summers, L. H. (2000). International financial crises: causes, prevention, and cures. The American Economic Review, 90(2), 1-16.

51. Taylor, J. B. (2009). The financial crisis and the policy responses: An

empirical analysis of what went wrong (No. w14631). National Bureau of

Economic Research.

52. Valickova, P., Havranek, T., & Horvath, R. (2015). Financial development and economic growth: A meta‐analysis. Journal of Economic Surveys, 29(3), 506-526.

53. Wachtel, P. (2017). Credit Deepening: Precursor to growth or crisis?.

54. Wang, J. (2006). Financial Liberalization in East Asia: Lessons from Financial Crises and the Chinese Experience of Controlled Liberalization.

Appendix 1

(38)

Germany Japan The UK

(39)

(40)

(41)

Appendix 2

Table A2.1.

Income Level Analysis of Model (1)

We regressed our logit model (1) with the samples split from the high-income country level line. Column 1 shows the model regressed for high-income countries and column 2 shows the model regressed for non-high income countries, which is every country other than high-income countries.

(1) (2) VARIABLES Crises High Income Crises Non-High Income Private Debt/GDP (%) 0.00279* 0.00402* (0.00163) (0.00224) Trade/GDP (%) -0.00740*** -0.00618*** (0.00215) (0.00198) Population Growth -0.0933 0.0610 (0.0679) (0.0549) Inflation 0.00223 0.000155* (0.00220) (8.95e-05) Observations 2,006 4,588 Number of country 52 131

Fixed Effect YES YES

(42)

Table A2.2.

Income Level Analysis of Model (2)

We regressed our logit model (2) with the samples split from the high-income country level line. Column 1 shows the model regressed for high-income countries and column 2 shows the model regressed for non-high income countries, which is every country other than high-income countries.

(1) (2) VARIABLES Crises High Income Crises Non-High Income Private Debt/GDP (%) T-1 0.00352** 0.00747*** (0.00163) (0.00226) Trade/GDP (%) T-1 -0.00527** -0.00703*** (0.00211) (0.00201) Population Growth T-1 0.0886 0.104* (0.0651) (0.0556) Inflation T-1 0.00454** 0.000133* (0.00219) (7.94e-05) Observations 2,006 4,588 Number of country 52 131

Fixed Effect YES YES

(43)

Appendix 3

Table A3.1.

Pre-2006 Regression of Model (1)

The results below are from regressing Model (1) to data samples from the period before the year 2006. Column 1 shows the regression without fixed effects while column 2 shows it with fixed effects.

(1) (2)

(44)

Table A3.2.

Pre-2006 Regression of Model (2)

Below are the results from regression Model (2) to data samples from the period before the year 2006. Just as we did before all independent variables has been lagged by one period. Column 1 shows the regression without fixed effects while column 2 shows it with fixed effects.

(1) (3)

(45)

Appendix 4

Table A4.1.

Below are the regressions of Model (1) and Model (2) using Liquid Liability/GDP as our main explanatory variable proxy-ing Financial Development. All other controls the same as our main logit regressions, which are Trade/GDP, Inflation and Population Growth. In Model (2) just as we did in our main regressions, we lagged the independent variables by one period.

(1) (2)

(46)

The Effect of Financial Development on the Likelihood of Economic Crises to Occur