The Effect of the Syrian Refugee Crisis on Lebanon’s External
Sector and the Credibility of Lebanon’s Peg to the US Dollar
MSc Thesis
Ali Smaili
11400595
University of Amsterdam Department of International Economics
Supervisor: Dr. D.J.M Veestraeten Second Reader: Prof. dr. Franc Klaassen
2
Statement of Originality
This document is written by Ali Smaili who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in producing it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
The Syrian refugee crisis that began in 2011 has been deemed the largest refugee crisis in recent times. With Lebanon being a close neighbor to Syria, the number of Syrian refugees entering Lebanon has reached a staggering 25% of the Lebanese population. The influx of refugees into Lebanon and the Syrian crisis has put serious implications on the Lebanese economy, especially the external sector and its influence on Lebanon’s peg to the US dollar. Therefore, the purpose of this research is to study the implications of the crisis on the credibility of this peg by examining the effects of the crisis on key components of Lebanon’s external sector. The following research constructs two Vector Auto-Regressions to estimate the effects of the Syrian crisis, and the influx of refugees, on Lebanon’s exports, imports, the Lebanese government’s external debt, foreign direct and portfolio investment, as well as foreign exchange reserves. The effects of the crisis on these components has allowed for the analysis of whether there may have been pressure on Lebanon’s currency, the Lebanese Lira. The results revealed that the influx of refugees, and the Syrian crisis as a whole, has increased the government’s external debt stocks, reduced exports, and influenced foreign exchange reserves held by the Lebanese central bank, whilst showing no effects of the crisis on imports, foreign direct and portfolio investment. Given the above results, the crisis may have increased the pressure on the Lira through increased external debt, lower exports, and the change of foreign exchange reserves. The results indicating that the crisis has not affected the capital inflows from abroad, should be seen as a glimmer of hope, since any reductions or reversals may diminish the credibility of the Lira’s peg to the US dollar and place Lebanon in an ultimately distressing position.
4
Table of Contents
1. Introduction ... 5 2. Brief Overview of the Government, the Economy of Lebanon and the Syrian Crisis ... 8 2.1. The Lebanese Government ... 8 2.1.1. Conflict and Division ... 8 2.1.2. Corruption ... 9 2.1.3. Government Debt and Fiscal Position ... 10 2.2. The Lebanese Economy ... 12 2.3. The Syrian Refugee Crisis and Lebanon ... 18 3. Literature Review on the Effects of Refugees on Host Countries ... 23 3.1. Qualitative Impacts of the Syrian Refugee Crisis on Host Countries ... 23 3.2. Vector Auto-Regression Models: Estimating the Impact of Refugees on Host Countries ... 24 4. Methodology and Data: Estimating the Effects of the Syrian Refugee Crisis on Lebanon’s External Sector ... 27 4.1. The VAR models ... 27 4.2.1. Variables ... 27 4.2.2. VAR model 1 ... 30 4.2.3. VAR model 2 ... 32 4.2.4. Stationarity and Lag Length ... 34 4.2.5. Goal of the VAR models ... 35 4.2. Data and Sample ... 35 5. Empirical Results ... 36 5.1. Stationarity Properties of the Variables ... 36 5.2. Lag Length ... 37 5.3. Dummy Variable Interpretation ... 38 5.4. Granger Causality Tests ... 39 5.5. Impulse Response Functions ... 43 5.6. General Overview of the Results ... 47 6. Conclusions ... 49 References ... 53 Appendices ... 545
1. Introduction
Ever since its 15-year civil war starting in 1975, Lebanon has been plagued by high levels of corruption, division, and conflict, which have all led to economic uncertainty in the country. The ongoing Syrian crisis that started in 2011, has affected all countries within the region, especially Lebanon, either through war contagion, trade, or indirectly through the influx of refugees. Lebanon, being one of the largest destinations for Syrian refugees (about 1.3 million), has been susceptible to various economic and political costs, painting a bleak image for the future of the Lebanese economy and particularly for its fixed exchange rate with the US dollar. Thus, the purpose of this thesis is to examine whether or not the influx of refugees into Lebanon may have put pressure on the Lebanese currency, the Lebanese Lira.
Lebanon has pegged the Lebanese Lira (L.L.) to the US dollar in 1999 at a rate of 1507 L.L./US$. Throughout this time period Lebanon has been experiencing recurrent current account deficits mainly due to the continuous accumulation of foreign debt and large trade deficits. However, these current account deficits have been more than offset by recurrent capital account surpluses and foreign reserve accumulation by the Banque du Liban (the Lebanese Central Bank), to support the pegged currency. With the Syrian crisis in its seventh year, Lebanon’s external sector and competitive position have been affected by the war environment neighboring Lebanon and the influx of refugees into Lebanon. These effects are pronounced through several channels, each threatening the credibility of Lebanon’s peg to the dollar.
Given that Lebanon and Syria are close trade partners, the Syrian war may definitely have disturbed Lebanese trade flows that either go to or through Syria. Since Lebanon has been experiencing large trade deficits throughout its history, a further aggravation of these deficits would increase the depreciation pressure on the Lira and damage the credibility of the peg. Furthermore, the influx of refugees into Lebanon has increased the financing needs of the
Lebanese government to support these refugees. With Lebanon being a country already burdened with high debt levels, it would seem reasonable for the Lebanese government to increase its borrowing from abroad in order to support these refugees. As a result, there would be an increase in demand for foreign currencies and would in turn increase the pressure on the Lira.
Another area where the Lebanese external sector may be affected is through the reduction or reversals of the capital inflows into Lebanon from abroad. In particular, the influx of refugees
6
into Lebanon may aggravate the tensions between the people of Lebanon and lead to internal conflicts, due to the opposing views towards Syria. This is, because the relationship between Lebanon and Syria has been rather unsettling ever since the start of the Lebanese civil war. These uncertain ties have been further aggravated following the assassination of Lebanon’s former prime minister Rafic Al-Hariri in 2005, where two movements in Lebanon were created, each having a different view on the involvement of Syria in this assassination. The possible internal conflicts in Lebanon, as a result of the influx of refugees, would signal instability of the Lebanese economy to the outside world, leading to probable reductions in capital inflows from abroad. Furthermore, the war environment neighboring Lebanon may also signal instability in Lebanon to the outside world, and could further result in reductions of investments from abroad. The probable reversals or reductions in capital inflows raises the most concern for the future of Lebanon. Reductions or reversals in capital inflows from abroad would definitely diminish the credibility of the peg, since the BDL would be induced to exhaust its foreign exchange reserves to counter-act the effects of these reversals on the Lira. If these reductions or reversals are so large that all foreign exchange reserves are exhausted, the collapse of the peg would be inevitable and would place Lebanon in a currency crisis, severely affecting its economy.
The motivation behind approaching this research question stems from the fear of Lebanon becoming more unstable than it already is. The ongoing Syrian war and the influx of refugees, reaching a staggering 25% of the Lebanese population, on top of Lebanon’s fragile democratic institutions and precarious economic and financial state, has placed Lebanon in the crosshair for further instability and possible collapse. The effects of the Syrian crisis on
Lebanon’s external sector and competitive position is important to analyze since any shocks to these areas could diminish the credibility of the peg and possibly place Lebanon in a political, economic, and financial crisis.
The methodology to be followed throughout this paper uses techniques of proven efficacy. A World Bank study in 2014 estimated the effect of the influx of refugees on
Lebanon’s trade using a Vector Auto-Regression (VAR) analysis, spanning January 2007 until August 2014. In this way, Granger causality testing and the construction of impulse response functions were performed to determine the influence of the influx of refugees on Lebanese trade. For this thesis, using the sample from January 2007 until April 2015, a VAR methodology similar to the one used in the World Bank study will be implemented to determine the effects of
7
the Syrian crisis on key components of Lebanon’s external sector. In particular, the effects on Lebanese exports, imports, the Lebanese government’s foreign debt, foreign direct investment, portfolio investment, and foreign exchange reserves will be examined. In this way, this study aims to take the previous research one step further by estimating the effects of the Syrian crisis on additional components of Lebanon’s external sector, whilst also allowing for the analysis of the implications on the credibility of the Lira’s peg to the dollar.
The remainder of this thesis is structured as follows. Chapter 2 presents the relevant background on the Lebanese government, economy, and the effects of the Syrian crisis to further motivate the above research question. Chapter 3 delivers a review of relevant empirical as well as qualitative work whilst also highlighting the intended contribution of this thesis. Chapter 4 gives a detailed explanation of the applied methodology, while chapter 5 provides the analysis of the results and their implications on the future of the Lebanese economy. Finally, Chapter 6 presents concluding remarks on the results coupled with policy advice for the government of Lebanon.
8
2. Brief Overview of the Government, the Economy of Lebanon and
the Syrian Crisis
Before introducing the methodology that will be applied in this study, a brief overview of the Lebanese government, economy, and the Syrian crisis will be given to provide the reader with the required background knowledge on the topic considered.
2.1. The Lebanese Government 2.1.1. Conflict and Division
For the past four decades, Lebanon has been fairly unstable, continuously “living close to the edge,” as the IMF stated in its 2015 country report (IMF, 2015). The 15-year civil war, starting in 1975, between the multiple religious groups and militias has plagued the country with continuous division, conflict and uncertainty that has yet to be resolved. Lebanon’s political system is a parliamentary system built upon a power-sharing agreement based on sectarian division. The Taif Accord of 1989, which ended the civil war, sought to resolve the conflict between the religious groups in Lebanon by equally appointing parliamentary seats between Christian and Muslim groups and legitimizing Syrian forces in Lebanon, which were present since 1976. The Taif accord may have ended the civil war, however division among the various Lebanese factions continued throughout the years that followed. These divisions led to the formation of two opposing political sides, namely the March 8 and March 14 movements. Many of the leading figures in both of these movements are civil war warlords and are also current members of the Lebanese parliament. These two groups were formed following the assassination of the former prime minister Rafic Al-Hariri in 2005, which the March 14 movement accused the Syrian regime of committing. As indicated above, Lebanon and Syria’s relationship has been complicated for decades. The Syrian crisis that began in 2011 heightened the already evident political disputes between the two political movements, further decreasing the legitimacy of the parliament as a decision-making body.
Decision-making by the parliament strongly depends on compromises that benefits all parties in the government. However, due to the division within the parliament, the needed compromises rarely occur. To put this in perspective, Lebanon lasted more than two years since
9
2014 without a president, where it took the parliament 46 rounds of elections to finally decide on an acting president. Furthermore, the waste management crisis in Lebanon that began in 2015 has yet to be resolved due to the inability of the parliament to come to a compromised decision on how to handle it. This further reveals the inefficiency of the government as a decision-making body, which raises concern on its ability to impose policies to stimulate the economy and absorb shocks. Because Lebanon has pegged the Lebanese Lira to the dollar at a rate of 1507 L.L./$ in 1999, monetary policy is subordinate as an effective tool to stimulate the economy, leaving fiscal policy as the main policy tool to do so and absorb any (external) shocks. However, due to the division existing in the government and the lack of cooperation, this policy tool has been rather subpar, exposing the Lebanese economy to detrimental internal and external shocks.
To make the above points clearer, Lebanon ranked at the 24th percentile on the government effectiveness index in 2014, 13th percentile on the rule of law index, and 3rd percentile on the political stability index (Moody’s Investors Service, 2016).
2.1.2. Corruption
In 2015 Lebanon ranked 123rd out of 177 countries on the corruption perception index, with 1 being the least corrupt (Moody’s Investors Service, 2016). The corruption exists in several areas throughout Lebanon. For example, vote-buying has been evident in Lebanon for years to secure certain seats in the government. Furthermore, according to the World Bank/IFC
Enterprise survey, 97.7% of surveyed companies admitted to paying bribes to secure a government contract. Also, according to a survey conducted by the Lebanese Transparency Association (LTA) in 2010, 74% of surveyed companies admitted to also bribing tax inspectors. Moreover, 65% of the companies were faced with a situation where they had to pay bribes to expedite certain government procedures, while 47.8% had to pay between 1% to 5% of their annual revenues as unofficial payments to government officials. This survey also revealed that 56% of companies had to pay around 1,000,000 L.L in addition to registration costs to register their businesses, while 18.2% of the companies had to pay between 4,000,000 and 5,000,000 L.L. In addition to the corruption in the business sector, the electricity sector in Lebanon has been plagued with corruption for decades, where 24-hour electricity is still not available to the population. According to the LTA survey, 31% of the surveyed companies were asked to pay bribes to the electricity sector in order to be supplied with electricity. The above corruption in
10
Lebanon is certainly a difficulty that new companies face when deciding whether to start up in Lebanon (Wickberg, 2012). With increased political tensions and a war environment neighboring Lebanon due to the Syrian crisis, the incentives to startup businesses in Lebanon could only diminish further.
2.1.3. Government Debt and Fiscal Position
Lebanon is known to have high levels of public debt as well as an inefficient tax
collection system which produces recurrent budget deficits. The figures below portray the fiscal position and amount of debt that the Lebanese government experiences.
11
12
Figure 1 indicates that the Lebanese government has been spending much more than it is earning in revenues from taxes. This has led to recurrent budget deficits as shown in Figure 2. Most alarming of the above graphs is Figure 3 which shows Lebanon’s government debt as a percentage of GDP. Ever since 1998, Lebanon’s public debt has exceeded 100% of GDP. In 2015 public debt reached 132% of GDP, where according to the IMF’s country report this number is “very high by international standards (IMF 2015, 5).” Furthermore, the report indicated that that the “need for fiscal adjustment is inescapable,” and that Lebanon “faces significant risks to public debt sustainability (IMF 2015, 14&37).” This is important to note, because, as stated in the previous chapters, the Syrian crisis has put a sizable strain on public services, amplifying the government’s need of financing. With the evident corruption, political instability, and high debt levels, this certainly raises concerns on Lebanon’s capability of dealing with the influx of refugees optimally.
The above discussed level of corruption, division, and conflict, that have led Lebanon into an unhealthy fiscal position, gives a general idea of how Lebanon has been plagued throughout the past few decades. These facts are important to note before moving forward, because the Syrian crisis has amplified the above political division and corruption. This significantly may dampen investor confidence in Lebanon, damaging a vital source of funding for Lebanon. The importance of investor confidence for Lebanon’s case will be further explained in the following section when the economic situation is discussed.
2.2. The Lebanese Economy
The destruction of infrastructure and the need for funding following the Lebanese civil war, which ended in 1990, made the current peg to the dollar a logical policy choice in 1999. Before the civil war, one US dollar was worth 3 L.L., where this value decreased substantially until 1992 when one US dollar was worth 2500 L.L. Without this peg, the Lebanese Lira would have experienced continuous depreciation following the civil war, due to Lebanon’s large trade deficits and the continuous borrowing from abroad to finance domestic infrastructure projects, amongst other things. Furthermore, the peg to the dollar was established to endure investor confidence in Lebanon, especially in Lebanese Treasury Bills, and attract capital inflows to fund the economy.
13
To further explain the economic situation in Lebanon throughout the past few decades, the below figures reveal Lebanon’s competitive position as well as the level of economic activity.
16
To begin with, Figure 8 shows real GDP growth in Lebanon since 1995. From 2007 until 2010 Lebanon experienced an annual GDP growth of around 9%. This may be attributed to the recovery following the war with Israel as well as Rafic Al Hariri’s assassination. However, this number plummeted following the start of the Syrian crisis, where GDP growth fell to about 2% in 2011 and has remained as is ever since. This indicates that the Syrian crisis may have affected economic activity within Lebanon, opening doors for further analysis of the effects of the crisis.
Figure 5 shows Lebanon’s trade balance from 1990 until 2015 as a percentage of its GDP. As depicted by the figure, ever since the end of the civil war in 1990, Lebanon’s trade deficit as a percentage of GDP increased dramatically. The figure indicates that Lebanon’s trade deficit is currently about 100% of its GDP. This is not surprising since Lebanon is a heavily importing country with a rather inefficient exporting sector. To make this point clearer, 80 percent of locally consumed goods in Lebanon are imported, revealing Lebanon’s damp
competitive position in international trade (Neaime, 2015). Furthermore, as stated in the previous chapters, Lebanon’s reliance on external financing has been evident ever since the end of the Lebanese civil war. Figure 7 depicts Lebanon’s total external debt stocks as a percentage of its GDP. Ever since 1993, Lebanon’s external debt stock as a percentage of GDP has been
increasing, reaching a staggering 110% in 2006. However, following the year 2006, the numbers have decreased substantially, but still remain relatively high at around 60% of GDP. As Lebanon continues to rely on external financing to this degree, the pressure on the current exchange rate peg may continue to rise, leaving the credibility of the peg vulnerable to external shocks.
As the trade balance and external debt accumulation are key components of the current account, figure 6 shows Lebanon’s current account balance as a percentage of its GDP. For the past few decades Lebanon has been running recurrent current account deficits, as depicted by the graph. These current account deficits amounted to 25% of GDP in 2010, where after the start of the Syrian crisis in 2011, the percentage increased further. A current account deficit of 25% of GDP is already relatively large, and with its increase following the Syrian crisis, the pressure on the current peg to the dollar may have well increased.
The recurrent trade deficits and external debt accumulation have led to recurrent current account deficits which may increase the depreciation pressure on the Lira. If the peg is not sustained, then the local-currency-value of this external debt outstanding will increase
17
However, these current account deficits have been more than offset by recurrent capital inflows and foreign exchange reserve accumulation by the Lebanese central bank, Banque du Liban (BDL). Figure 10 below reveals BDL’s foreign exchange reserve accumulation throughout the past few decades.
As shown in the above figure, Lebanon’s central bank has been successful in accumulating foreign exchange reserves throughout the years to sustain the peg, where these reserves amounted to 39 billion US$ in 2014.
The capital inflows have been mainly attributed to deposit inflows into the banking sector from residents and investments in Lebanese Treasury Bills. These deposits from
non-residents, 23 percent of total deposits, are crucial in allowing Lebanese domestic banks to
channel these funds to the private and public sector, with the goal of financing Lebanon’s current account and budget deficits (IMF, 2015). Even though this dependency on the banking sector to finance Lebanon’s current account and budget deficits has been successful in the past, it poses a
18
threat for the near future. Specifically, given that the Syrian crisis may dampen investor confidence in Lebanon, these deposits may be reduced or even withdrawn. This not only increases the pressure on the Lira to depreciate, but would also oblige the central bank to sell foreign reserves and sustain the peg to the dollar.
This brief overview of the political and economic situation in Lebanon reveals that even before the Syrian crisis, Lebanon had already been exposed to continuous risk, mainly due to the political instability within the country. This brings us to a critical point in time where the future of Lebanon has never been more uncertain, particularly with the ongoing Syrian crisis.
2.3. The Syrian Refugee Crisis and Lebanon
Given that Lebanon is a close neighbor of Syria, it has become one of the largest destinations for refugees since the start of the Syrian crisis. According to the United Nations High Commissioner for Refugees (UNHCR) around 1.2 million refugees were documented in Lebanon in 2015, amounting to a quarter of the entire Lebanese population. However, due to the difficulty in documenting all refugees entering Lebanon, the Lebanese government stated that the number of Syrian refugees has surpassed the 1.5 million mark. Figure 11 shows the number of Syrian Refugees in Lebanon since 2005.
19
The large number of Syrian refugees entering Lebanon has put a sizable strain on public services such as healthcare, education, water, utility, etc., increasing the amount of government spending needed to accommodate these refugees. This raises concern, because of Lebanon’s already tainted fiscal position and political instability. The World Bank estimated that the fiscal strain over 2012-2014 amounted to 2.6 billion dollars (5.5 percent of GDP) and that it would cost an additional 2.5 billion dollars to restore the public services system to pre-crisis levels (IMF, 2015). Furthermore, the Lebanese Crisis Response Plan (LCRP) estimated that funding requirements for 2016 amounted to about 2.5 billion dollars (4.7 percent of GDP), up from around 2 billion dollars in 2015 (Shellito, 2016). However, Lebanon has been subject to aid from foreign organizations and countries since the start of the crisis. Yet, according to the below figure, the aid that Lebanon has received each year has not reached the estimated required funds to accommodate the refugees.
20
The above figure, extracted from the Lebanon Crisis Response Plan for 2017-2020, reveals that for every year after the start of the Syrian crisis, the aid received by Lebanon has not covered the required funding (UN, 2017). This indicates that the Lebanese government would need to
increase its spending to continue to accommodate the refugees. However, since Lebanon is already burdened with high levels of debt, increased borrowing to fund the refugees will
probably occur. The increased borrowing and spending by the government will further aggravate Lebanon’s public and external debt, intensifying the dependency on capital inflows from abroad. This brings us back to the channel described previously, i.e. whether these capital inflows from abroad will be reduced or reversed due to the current war climate neighboring Lebanon. Since the dependency on these capital inflows, to finance the needed government spending, has increased substantially post-crisis, the risk that Lebanon faces in the near future has increased significantly. Figure 12: Foreign Aid to Lebanon Since the Start of the Syrian Crisis Source: United Nations Lebanon Crisis Response Plan 2017-2020 Note: RRP5 and RRP6 represent the donors for the Syrian Regional Response Plane in the years 2013 and 2014 respectively. LCRP represents the donors for the Lebanon Crisis Response Plan.
21
The Syrian crisis that began in 2011 has deepened the problems that Lebanon has been facing, and successfully avoiding, for the past few decades. In particular, this crisis has not only affected the Lebanese economy directly, but also indirectly by aggravating the already evident internal and external political tensions. This could further dampen investor confidence in Lebanon and affect the capital inflows that have been crucial in sustaining the currency peg to the dollar. This is highly probable, given that similar instances have occurred in the past. For example, the collapse of the Lebanese government in 2011 led to deposit outflows amounting to 1.1 billion US dollars, while 2 and 3 billion US dollars were withdrawn following the
assassination of the prime minister of Lebanon, Rafic Hariri, and the 2006 war respectively (Hentov et al, 2012). This by itself raises concern for the sustainability of the current peg, due to the high dependency on these deposits in the banking system to finance Lebanon’s current account and budget deficits.
Not only is it probable that the Syrian refugee crisis has increased the pressure on the Lira through the channel of increasing political turmoil, but also through multiple economic channels. In particular, economic growth has severely slowed down following the start of the crisis. After an average growth rate of 8 percent per year from 2007 until 2010, growth has slowed to 1.7 percent in 2011. This can also be attributed to a decline in tourism, a key sector that has fueled the economy of Lebanon for years through increased domestic spending and inflows of foreign currencies. Since 2010, the number of tourists visiting Lebanon has declined by 30 percent, worsening the already weak competitive position of Lebanon (Moody’s Investors Service, 2016). Moreover, Lebanese exports have taken a hit following the Syrian crisis due to the close link of the two economies. About 6 percent of exports go directly to Syria and about 20 percent through Syria to several neighboring countries (Hentov et al, 2012). This close link enlarged the effect of the crisis on the Lebanese exports, which fell about 31 percent from 2010 until 2015, further aggravating Lebanon’s already evident trade and current account deficits, and further increasing the probable pressure on the Lira (Moody’s Investors Service, 2016). As shown in figure 6, Lebanon’s current account deficits as a percentage of GDP increased substantially following the start of the crisis in 2011.
Examining whether there have been any signs of probable pressure on the Lira following the Syrian crisis is important for Lebanon’s future. Even though the central bank has been successful in sustaining the peg to the dollar by accumulating high levels of foreign reserves,
22
external shocks of this magnitude may limit BDL’s ability to sustain this peg. If capital inflows from abroad are reversed or reduced to a large extent, due to the tainted investment climate, the BDL may not be able to withstand these large reversals. Also, given that Lebanon’s current account balances have increased following the start of the crisis, the central bank is in an even more distressing position.
23
3. Literature Review on the Effects of Refugees on Host Countries
Given that the Syrian Crisis is a relatively recent event in history, the literature considering its impact is limited. Furthermore, most of the literature surrounding this topic is based on the humanitarian rather than the economic implications of the Syrian crisis. However, in this chapter we present the key academic papers that relate to the research question of this thesis. The following chapter is divided into three sections. In the first section, we present the literature surrounding the qualitative economic impact of the Syrian refugee crises on the countries hosting these refugees. Section two provides the key literature focusing on the quantitative research that estimates the impacts of the influx of refugees on the countries that host them using the VAR methodology. The literature provided in this second section builds the framework for the methodology to be followed in the empirical analysis, which will be presented in chapter 4. Finally, section 3 presents an interesting paper that will help motivate why this research question has indeed been tackled.
3.1. Qualitative Impacts of the Syrian Refugee Crisis on Host Countries
To begin with, an IMF study by Aiyar et al. (2016) examined the economic impact of Syrian refugees on host countries. The authors concluded that the influx of refugees strains the services systems in the host countries and substantially increases government expenditures on housing, water, food, health care, education, and utilities. They estimated that the “average budgetary expenses for asylum seekers in EU countries could increase by 0.05 and 0.1 percent of GDP, respectively.” However, they indicated that these effects may vary from one country to another, depending on the structure of the host country’s economy. Thus, the impact of the influx of refugees for the Lebanese case may differ substantially, due to the fact that Lebanon is less developed than most, if not all, EU countries. Furthermore, Shellito (2016) qualitatively examined the economic effects of refugees on host countries, in general, and for the Lebanese case with the influx of Syrian refugees in particular. With about 1.3 million refugees registered in Lebanon (20% of the Lebanese population), Shellito (2016) indicates that the influx of Syrian refugees in Lebanon has put a strain on public services, particularly education, health care, and waste management. According to a 2013 World Bank study, Shellito notes that these strains have cut the annual GDP growth in Lebanon by 3 percent, doubled unemployment and costed the government about 3 billion US dollars. Given this qualitative analysis by Shellito, one can
24
observe that the Syrian crisis has affected the Lebanese economy, especially through increased government spending. The above research papers indicate that the influx of refugees increases the financing needs of the respective governments that host them. As shown in the previous chapter, the Lebanese government is already burdened with high levels of internal and external debt. Thus, estimating the effect of the influx of refugees into Lebanon on the Lebanese
government’s external debt is feasible given the above literature. If the influx of refugees into Lebanon increases the Lebanese government’s external debt, then increased demand for foreign currencies results, which puts pressure on the Lira to depreciate.
3.2. Vector Auto-Regression Models: Estimating the Impact of Refugees on Host Countries
The complexity of modeling the impact of refugees on the economies of host countries using VAR analysis, stems from the question of whether the influx of refugees is best included as endogenous or exogenous to the model in question. Including the influx of refugees as an endogenous variable may be shocking since it is difficult to explain how economic variables used in the model influence the level of refugees entering the country. This may lead one to prefer including the influx of refugees as an exogenous variable. However, including the
refugees as exogenous to the system comes at a certain cost. The use of Granger causality testing and the formation of impulse response functions (IRF) are no longer applicable in that case. To be more concrete, Granger causality testing and the creation of IRFs only include the
endogenous variables in question. Thus, if the influx of Syrian refugees is included as
exogenous, the dynamic relationships between the refugees and the other variables cannot be analyzed using the above tests. Most of the literature surrounding this topic includes the influx of refugees as endogenous to the system to allow for the above Granger causality testing and the formation of IRFs. The literature justifies including the refugee variable as endogenous by indicating that economic variables that reveal the health of an economy may incline refugees to seek sanctuary in the corresponding country.
To begin with, Fakih et al. (2016) estimate the effect of the influx of refugees, resulting from the Syrian crisis, on the labor market in Jordan. In their study, they model these effects by constructing three VAR models, each including three variables. Each VAR model includes the
25
influx of refugees and an indicator for the level of economic activity as endogenous to the system, where the third variable differs in each model. The third variable for each model captures the labor market activity within Jordan. Specifically, these variables are employment rates, the unemployment rate, and the labor force participation rate. Using quarterly data covering the period from 2007Q4 until 2014Q4, the results of their estimations reveal no Granger causality running from the influx of refugees to the labor market variables indicated above. Furthermore, the IRFs also reveal no impact of a positive shock to refugees in Jordan on the labor market variables. Since the impact of the influx of refugees on Lebanon’s labor market is not the focus of this study, the use of this study for this thesis may be confusing. However, this study provides a sufficient background on how to model the effect of the influx of refugees on the economy of their respective host countries using VAR analysis.
A more relevant report to the research in question is a World Bank study by Calì et al. (2014). In their study, they estimate the effect of the influx of Syrian refugees into Lebanon on several balance of payments (BOP) components for Lebanon. In particular, using monthly data ranging from January 2009 to August 2014, they estimate seven VAR models, each including three endogenous variables. For each VAR model, the influx of refugees and a coincident indicator constructed by the World Bank for Lebanon are included as endogenous, where the third variable differs in each model. The third endogenous variable in the VAR models consist of BoP components such as exports of services, imports of food, imports of capital goods, imports of consumption goods, inflows of remittances, outflows of remittances, and BoP errors and omissions. Furthermore, each model includes a set of exogenous variables such as the total number of refugees in the countries neighboring Lebanon, dummy variables to proxy for the security situations in Lebanon such as incidents of car explosions and fighting in Tripoli, and a dummy variable to account for the divergences in the BDL’s BoP data from other sources, evident after 2011. The results of their estimations reveal no significant Granger causality running from the influx of refugees to any of the BoP variables. However, their IRFs indicate that a positive shock to the refugees in Lebanon significantly increases exports of services, decreases the imports of capital, and reduces the BoP errors and omissions. They indicate that the inability to obtain stronger results may be due to the quality of the data for Lebanon.
Specifically, they reveal that historical data is subject to large revisions and, as stated previously, the data obtained from the BDL diverges from other sources. Thus, the above study not only
26
provides a sufficient framework for modeling the influx of refugees into Lebanon, but also provides evidence that the influx of refugees has affected several of Lebanon’s BoP components. This provides room for the estimation of the influx of refugees into Lebanon on other variables of Lebanon’s external sector, which is what this thesis will attempt to tackle.
3.3. Currency Crises Through Sovereign Defaults
As stated before, the Lebanese government has been accumulating large amounts of external debt throughout its history. According to the above literature examined, the Syrian refugee crisis has increased the financing needs of the Lebanese government to support the influx of refugees into Lebanon. Since Lebanon is already burdened with high levels of external debt, further borrowing from abroad to finance the refugee’s needs seems plausible. If this indeed is the case, increased borrowing from abroad may well increase the depreciation pressure on the Lira, further diminishing the credibility of the peg. Thus, the fear of Lebanon defaulting on its external debt is heightened with the current influx of refugees. One academic paper that provides a sufficient background for the above fear is a study conducted by Reinhart (2002). In her study, Reinhart examined the relationship between foreign debt and exchange rate crises in both developing and developed countries. In her sample of 59 countries throughout the years 1970-1999, she found that about 84% of all sovereign debt default events were followed, within two years, by currency crises. The above results obtained by Reinhart provide sufficient motivation to why this research question in particular is important for the future of the Lebanese economy.
This thesis will estimate the effect of the influx of refugees into Lebanon, and thus the Syrian crisis, on several components of Lebanon’s external sector. Since the study by the World Bank estimated the effects on Lebanese trade, this thesis will go one step further. With an increased sample size compared to the one used by the World Bank, this thesis will attempt to estimate the impact of the influx of refugees, and the Syrian crisis as a whole, on Lebanon’s exports and imports, foreign direct investment within Lebanon, portfolio investment, foreign exchange reserves held by the BDL, and the amount of external debt of the Lebanese
government. By doing this, potential implications on the value of the Lira will be discussed.
27
4. Methodology and Data: Estimating the Effects of the Syrian
Refugee Crisis on Lebanon’s External Sector
The first section of this chapter will present the construction of the VAR models that will be used to analyze whether the Syrian refugee crisis may have put pressure on the Lebanese Lira. The VAR models will allow for the examination of the dynamic relationship between the influx of refugees and the economic variables, whilst also allowing for Granger causality tests and the construction of impulse response functions. After presenting the methodology that will be undertaken, section 2 provides details on the data that will be used to proceed with the below study.
4.1. The VAR models
To examine whether the Syrian refugee crisis may have applied pressure on the Lebanese Lira, two vector auto-regression (VAR) models will be constructed. The use of VAR
methodology is useful for this study in particular. Specifically, it allows for the examination of effects between variables that need not have any structural relationship. Furthermore, it allows for the construction of impulse response functions to determine the effect of a shock of one variable on another. This is definitely appropriate for this study, because the influx of Syrian refugees can be treated as a shock. Before introducing the two VAR models, the variables that will be used are introduced below.
4.2.1. Variables
Public External debt (𝒑𝒆𝒅𝒕): Since the Syrian refugee crisis has put a sizable strain on public services and has increased the government’s need of financing, the use of public external debt is thus valid for the analysis that follows, assuming that the government finances the needs of the refugees by borrowing from abroad. This assumption is indeed reasonable since, as shown in the previous chapter, one way that the Lebanese government has been raising funds for spending throughout the past few decades is by borrowing from abroad. Furthermore, since the foreign aid from international organizations, as shown in the previous chapter, has not been sufficient to cover the refugee needs and that Lebanon is already burdened with large amounts of debt, further borrowing by the government from abroad is thus reasonable. The continuous accumulation of
28
external debt puts pressure on the exchange rate to depreciate because of the increased demand for foreign currencies and the aggravation of the current account deficits. Thus, by examining the effect of the Syrian refugee crisis on the government’s external debt, the effect on the Lira will be analyzed, in turn.
Exports (𝒙𝒕): In the previous chapter, it was indicated that a considerable amount of Lebanon’s exports went directly to Syria or indirectly through Syria. Furthermore, tourism in Lebanon decreased by 30% after 2010, which could be due to the influx of refugees and war environment neighboring Lebanon. If exports are reduced, then the Lira would experience depreciation pressure due to the aggravation of Lebanon’s trade deficits and current account deficits. Thus, examining the effect of the Syrian refugee crisis on Lebanon’s exports will in turn allow for the analysis of whether the crisis may have increased the pressure on the Lira.
Imports (𝒎𝒕): Given that Lebanon imports 80% of locally consumed goods, and Syrian refugees make up around 25% of Lebanon’s population, an increase in imports following the start of the Syrian crisis can be expected. If imports were to increase, then additional
depreciation pressure is put on the Lira, due to the further aggravation of Lebanon’s trade and current account deficits.
Foreign exchange reserves (𝒇𝒆𝒓𝒕): As stated in the previous chapters, the Lebanese central bank has sustained the peg to the dollar by accumulating high levels of foreign exchange reserves. Therefore, it is vital to examine the effect of the Syrian refugee crisis on the
accumulation of foreign exchange reserves, because any change in foreign exchange reserves implies pressure on the Lira.
Foreign Direct Investment (𝒇𝒅𝒊𝒕) and Portfolio Investment (𝒑𝒊𝒕): Foreign direct investment inflows and portfolio investment liabilities for Lebanon will be included to further examine the effect of the Syrian crisis on the capital inflows that have been key in keeping a balance of payments equilibrium, and in turn, a sustainable fixed exchange rate. The use of portfolio investment liabilities instead of both portfolio investment assets and liabilities is important to indicate since the liability side is what determines whether foreigners are investing in Lebanon.
29
Portfolio Investment is used to encompass all investments in Lebanon by foreigners, including investments in stocks, bonds and certificate deposits (CDs) in Lebanese banks. As stated in the previous chapter, investments in Lebanese treasury bills have been a key source of funding for the government. Thus, the inclusion of this variable is valid for the following study.
The reason for including the above economic variables should be evident. External debt
accumulation and the trade deficit make up the current account, while foreign exchange reserves, foreign direct investment, and portfolio investment are key components in the capital account. Thus, examining the effect of the Syrian Crisis on the above variables would allow for the analysis of whether the Syrian refugee crisis may have well put pressure on the Lira.
Coincident Indicator (𝒄𝒊𝒕): This variable is included to capture and control for the current state of economic activity within Lebanon. The use of this indicator instead of GDP itself is due to the lack of monthly data for GDP. The World Bank recently constructed a coincident indicator for Lebanon, however the data for this ends in December of 2013. The BDL provides sufficient monthly data for a coincident indicator for Lebanon. Thus, the use of the variable provided by BDL will be used to control for the general economic activity within Lebanon.
The number of Syrian refugees in Lebanon (𝒓𝒆𝒇𝒕): Including the number of Syrian refugees in Lebanon will help in determining whether the influx of Syrian refugees has influenced the above economic variables and may have put pressure on the Lira.
Dummy variable (𝑫𝒕): The dummy variable will allow for the analysis of whether the above economic variables have changed, on average, after the start of the Syrian crisis. This variable takes the form: 𝐷1 = 0, 𝑡 ∈ 𝐽𝑎𝑛𝑢𝑎𝑟𝑦 2007 − 𝑀𝑎𝑟𝑐ℎ 2011 1, 𝑡 ∈ 𝐴𝑝𝑟𝑖𝑙 2011 − 𝐴𝑝𝑟𝑖𝑙 2015
All of the above variables will be used in level form in the below regressions since any other transformation of the variables would not be appropriate. For example, taking percent changes or the log operator would not be reasonable since some of the above variables take on negative and
30
zero values. Moreover, the above economic variables, except the coincident indicator, are used in millions of US dollars.
4.2.2. VAR model 1
The first VAR model that will be estimated takes the form: 𝑌1= 𝛼 + 𝛽M𝑌1NM
O
MPQ
+ 𝛾1𝐷1+ 𝜀1 Where 𝑌1 is a vector of five endogenous variables that includes:
• Foreign direct investment 𝑓𝑑𝑖1
• Portfolio Investment (𝑝𝑖1) • Foreign exchange reserves (𝑓𝑒𝑟1)
• The number of Syrian refugees in Lebanon (𝑟𝑒𝑓1) • Coincident Indicator (𝑐𝑖1)
𝛼 and 𝛽 are 5𝑥5 matrices including the coefficients, 𝐷1 is the dummy variable, 𝛾1 is a 5𝑥1 vector of coefficients, 𝜌 is the lag length, and 𝜀1 is a 5𝑥1 vector of error terms.
By estimating the above model, we can determine whether the Syrian crisis has influenced the levels of foreign direct investment, portfolio investment, and foreign exchange reserves held by the BDL, through the Granger causality test and by examining the coefficient of the time dummy variable. This would indirectly allow for the analysis of whether there may have been pressure on the Lira following the start of the Syrian crisis. If the results of the above model indicate, for example, that the number of Syrian refugees Granger cause FDI inflows in such a way that past values of Syrian refugees lower current FDI inflows, then one may come to the conclusion that the Syrian crisis may have increased the depreciation pressure on the Lira. This may certainly be the case, because the influx of refugees would dampen investor confidence in Lebanon due to their effects on the economy. For example, as stated before, the influx of refugees has strained public services in Lebanon due to the increased need of financing to support these refugees. However, since Lebanon is already overwhelmed with high levels of public debt, efficiently allocating resources to accommodate the refugees will be difficult, where this may severely affect the economy of Lebanon and investor confidence, in turn. Also, the increased need of financing may result in an increase in taxation, which will further reduce
31
incentives to invest in Lebanon. Furthermore, the influx of Syrian refugees may increase internal conflict within Lebanon. As stated before, Lebanon’s parliament is split into two political
movements, namely the March 8th and 14th movements. While the March 8th movement supports
the Assad regime in Syria, the March 14th movement opposes it. Thus, the influx of Syrian refugees may stir up civil conflicts within Lebanon due to the two opposing views on the conflict in Syria. This may further dampen investor confidence in Lebanon, and support the hypothesis that the influx of Syrian refugees will reduce the levels of foreign investment in Lebanon. The above arguments also apply to portfolio investment, where the number of Syrian refugees will weaken investor confidence in Lebanon and lead to a reduction in portfolio investment, a key source of funding for the economy of Lebanon, especially with investments in Lebanese treasury bills. Thus, if the results of the above model indicate that past values of Syrian refugees in Lebanon help predict lower values for FDI and portfolio investment, then the Syrian crisis may have reasonably increased the depreciation pressure on the Lira.
Testing whether the influx of Syrian refugees Granger causes foreign exchange reserves also helps in determining whether the crisis could have increased the pressure on the Lira. For example, if past values of refugees increase the amount of foreign exchange reserves held by the Bank du Liban, then one may come to the conclusion that the crisis may have increased the pressure on the Lebanese Lira. This may indeed be the case, because BDL may see the influx of refugees as a potential shock to the economy, and in this way, they may increase its foreign reserve buffer to counter-act any potential reduction in capital inflows from lower investor confidence.
The inclusion of the number of Syrian refugees as endogenous to the system may be puzzling, since it will be difficult to analyze whether the economic variables influence the number of refugees. Although, this could be the case, where a healthy economy may incline refugees to seek refuge in the country under consideration. However, given Lebanon’s unstable economy and that it is a close neighbor of Syria, this link may not be reasonable. Nevertheless, these links will be disregarded since it is not the focus of this study. Furthermore, including the number of refugees as endogenous to the system allows for the construction of impulse response functions. In this way, a graphical representation of how a shock in the number of refugees affects the above economic variables will be provided for further analysis. As stated before, the
32
coincident indicator is included to control for economic activity, where the exclusion of this variable may impair the specification of the model.
The inclusion of the dummy variable will help in determining whether these economic variables changed, on average, after the start of the Syrian crisis. This will further allow for the analysis of whether the Syrian refugee crisis may have increased the pressure on the value of the Lira. For example, if the coefficient of the dummy variable turned out to be significantly
negative for portfolio investment, for example, then on average portfolio investment was lower after the start of the Syrian crisis. This will further support the argument that the Syrian crisis could have increased the depreciation pressure on the Lira. It is important to include this dummy variable, because the Syrian crisis can be seen as a structural break/shock for the economy of Lebanon. Unlike the inclusion of the number of refugees, the time dummy variable would allow for the comparison of the trends of the above variables before and after the start of the Syrian crisis.
The impulse response functions will be generated based on the Cholesky decomposition identification. The ordering specification will be 𝑟𝑒𝑓, 𝑐𝑖, 𝑓𝑑𝑖, 𝑝𝑖, 𝑓𝑒𝑟 . By doing this, each endogenous variable affects the variable to the right of it. We include the number of Syrian refugees first, because the focus of this study is to determine the impact of the influx of refugees on the above economic variables. In this way, the IRFs reveal the impact of a positive shock to the number of refugees on the level of economic activity, FDI, portfolio investment, and foreign exchange reserves. This ordering is logical since placing the refugee variable in any other position would not be reasonable, since it would be hard to argue that the above economic variables influence the number of refugees in Lebanon.
4.2.3. VAR model 2
The second VAR model that will be estimated is as follows: 𝑋1 = 𝛼 + 𝛽M𝑋1NM
O
MPQ
+ 𝛾𝐷1+ 𝜀1 Where 𝑋1 is a vector of four endogenous variables that includes:
33
• Exports 𝑥1 • Imports (𝑚1)
• The number of Syrian refugees in Lebanon (𝑟𝑒𝑓1)
• Coincident Indicator (𝑐𝑖1)
The other vectors and matrices are similar to the ones in model 1.
Estimating this model will allow for determining whether the influx of Syrian refugees influences public external debt, exports, and imports. The inclusion of these variables as
endogenous will allow for the analysis of whether the Syrian refugee crisis has affected the key components that make up the current account. If the results of the above model indicate that past values of refugees influence current values of external debt, for example, in such a way that they increase the amount of external debt, then one may come to the conclusion that the Syrian refugee crisis may have increased the depreciation pressure on the Lira. This hypothesis is logical, because as stated before the government’s need for financing has increased with the influx of Syrian refugees. Since Lebanon’s government has depended on foreign borrowing to finance domestic spending, the influx of refugees may have increased the amount of external debt to accommodate the refugees, and in turn, increased the depreciation pressure on the Lira.
Another way the above model will help in determining whether the Syrian crisis has increased the depreciation pressure on the lira is through the effect on Lebanon’s trade balance. If the results indicate that the influx of Syrian refugees Granger causes exports and imports in a way that it increases Lebanon’s imports and decreases its exports, then the conclusion of probable increased depreciation pressure will be satisfied. This could indeed be the case since past values of refugees may influence the amount of tourism in Lebanon, a key source of export revenue that has fueled the economy of Lebanon for years. As stated before, tourism in Lebanon decreased by 30% after 2010. The intuition behind this could be that an influx in refugees would increase the political and civil tensions within Lebanon substantially, reducing the amount of tourism in Lebanon. This would increase the depreciation pressure on the Lira, because of the decrease in demand for it as well as the aggravation of Lebanon’s trade deficits. Furthermore, the influx of refugees may further aggravate Lebanon’s trade deficits through an increase in imports. Since Lebanon imports 80% of locally consumed goods, and Syrian refugees make up around 25% of Lebanon’s population, an increase in imports would be plausible. This would also
34
indicate depreciation pressure on the Lira through increased demand of foreign currencies and further aggravation of Lebanon’s trade deficits.
Similar to the first VAR model, testing whether the above economic variables Granger cause the number of refugees will be disregarded for similar reasons.
The impulse response functions for this model will also be based on the Cholesky decomposition identification. The ordering specification in this case will be 𝑟𝑒𝑓, 𝑐𝑖, 𝑥, 𝑚, 𝑝𝑒𝑑 . The reason for this ordering is similar to the one in model 1, because any other ordering may not be logical.
Each economic variable above is graphed with the number of Syrian refugees to portray any relation between the variables. The graphs are found in Appendix A. The graph for foreign exchange reserves reveals that reserves have slightly increased following the influx of refugees. Furthermore, public external debt seems to have increased following the influx of refugees, whilst exports appear to have fallen. The other graphs in appendix A do not seem to show any clear relationship with the influx of refugees.
4.2.4. Stationarity and Lag Length
Before proceeding with the estimations, the above variables will be tested for a unit root using the Augmented Dickey-Fuller (ADF) test. If any of the variables is non-stationary (i.e. contains a unit root or is 𝐼(1)), then we apply the same test to the first difference of the variable to determine whether it is integrated of order two. All variables integrated of order one will be differenced once, while the variables integrated of order two would be differenced twice. In this way, all variables used in the above regressions are stationary and spurious regressions are avoided.
The optimal lag length (𝜌) will be chosen to pass the test for no serial correlation in the residuals, the normality test, and the test for heteroscedasticity. The Breusch-Godfrey Lagrange multiplier (LM) test will be used to test for serial correlation, while the Jarque-Bera test will be used to determine whether the data is approximately normally distributed. Furthermore, White’s heteroscedasticity test will be used to ensure homoscedasticity across error terms.
35
4.2.5. Goal of the VAR models
The goal of estimating the above models, and the thesis as a whole, is not to determine whether the current peg to the dollar will collapse due to the Syrian crisis. Instead, the goal of this thesis is to determine whether there have been any reasonable signs of pressure on the Lira, given the current environment with the Syrian crisis.
4.2. Data and Sample
The data used throughout this study are extracted from the IMF’s database and Banque du Liban’s website. The data for Syrian refugees is extracted from the UNHCR database and are considered as “total persons of concern” by the UNHCR. These numbers consist of refugees and “asylum-seekers,” people in refugee-like conditions, internally-displaced people (IDPs), stateless people, and “other persons of concern to the UNHCR.” For simplicity, we refer to “total persons of concern” as the total number of Syrian refugees.
The UNHCR indicates that after May 6th of 2015, Lebanon’s government suspended any new registration of refugees. The refugees awaiting registration after this date are no longer included in the total refugee count. According to the data, the total number of Syrian refugees in Lebanon began to fall substantially following this suspension. This indicates that the refugees awaiting registration, that were included prior to May 6th of 2015, are no longer included in the data. This understates the actual number of refugees in Lebanon after this date, and thus may not allow for the accurate estimation of the relationships between the refugees and the economic variables under consideration. Due to this limitation in the data, the sample period used in the above study will be from January 2007 until April 2015. In this way, the limitations on the data for the Syrian refugees in Lebanon would not limit the accuracy of the results obtained in the next chapter. Furthermore, this sample period would also account for the relationships between the economic variables prior to the Syrian crisis, when the number of refugees was zero. Then we are able to examine the effects of the influx of Syrian refugees.
Other data caveats to note are that some of the Balance of Payments data obtained from the BDL diverge from other sources after 2011. Furthermore, the BDL indicates that their historical data is susceptible to large and frequent revisions. The above limitations to the data may diminish the quality of the results obtained.
36
5. Empirical Results
The following chapter is structured as follows. Section 1 examines the stationarity properties of the variables in question to determine their order of integration. Section 2 provides the results of the LM test for serial correlation as well as White’s heteroscedasticity test to determine the optimal lag length for each VAR model. Furthermore, sections 3, 4, and 5, present the results for the dummy variable coefficients, the Granger causality tests, and impulse response functions, respectively. This chapter concludes with an overall summary of the empirical results.
5.1. Stationarity Properties of the Variables
We begin by investigating the stationarity properties of the variables used in the VAR models. The Augmented Dickey-Fuller (ADF) test is used to determine whether a variable contains a unit root. The null hypothesis of this test states that the series contains a unit root, i.e. is non-stationary. If the null hypothesis is accepted for the variable in question, the test will also be applied to the first difference of the variable to determine whether it is integrated of order two. The first difference operator will be applied to all 𝐼(1) variables to remove the unit root, whereas for any 𝐼(2)variables, the second difference operator will be applied. In this way, all variables used in the VAR models are stationary and spurious regressions are avoided.
For the ADF test, we choose three specifications: with an intercept, with an intercept and time trend, or with neither. The choice of which specification to use is explained in Appendix B, where the detailed results of the above tests are found. The below table presents a summary of the ADF test results for each variable.
37
Lag Length ADF value p-value Order of
Integration 𝑟𝑒𝑓 2 −1.579641 0.7940 𝐼(1) 𝑐𝑖 0 −3.869639** 0.0169 𝐼(0) 𝑝𝑒𝑑 1 −1.261373 0.8913 𝐼 1 𝑥 0 −3.937233*** 0.0026 𝐼(0) 𝑚 3 −2.347543 0.1595 𝐼 1 𝑓𝑒𝑟 0 −1.520049 0.5195 𝐼 1 𝑓𝑑𝑖 5 −2.968680 0.1466 𝐼(1) 𝑝𝑖 0 −7.769509*** 0.0000 𝐼 0
Notes: The lag length is determined automatically based on the Schwarz Information Criterion (SIC). *, **, and *** denote
rejection of the null hypothesis at the 10%, 5%, and 1% level, respectively.
According to the above results, 𝑟𝑒𝑓, 𝑝𝑒𝑑, 𝑚, 𝑓𝑑𝑖, and 𝑓𝑒𝑟 are integrated of order one. Thus, the first difference of these variables will be used in the VAR models. The first difference of the above variables will be denoted with the letter “d” preceding their name.
5.2. Lag Length
Choosing the optimal lag length for time series data comes with a certain trade-off. On the one hand, increasing the lag length would remove the serial correlation in the error term. This is important to take into account, because if the error terms are correlated with one another, then the usual OLS standard errors and t-statistics are no longer valid. However, the more lags included, the lower the degrees of freedom available for estimation, where this may lower the accuracy of the results. Thus, the lowest lag length that passes the residual tests for no serial correlation, no heteroscedasticity, and normality will be used. Taking account of
heteroscedasticity and normality in the residuals is also key, because heteroscedastic or non-normal residuals could lead to invalid t-statistics and standard errors for inference.
At no lag-length did any of the VARs pass the residual test for normality. As a second-best criterion, the lowest lag-length to pass the residual tests for no serial correlation and homoscedasticity will be used.
38
The LM test requires us to choose a lag length and reports the serial correlation in the residuals at each lag. According to the literature, if there is no serial correlation at the first lag then the conclusion of no serial correlation is accepted. However, to be on the safe side, the lag length that passes the test at the first lag and more than half of the other lags will be chosen. For both VAR models, the optimal lag length turned out to be 5. The results of the LM test and White’s heteroscedasticity test for each VAR model at the chosen lag lengths can be found in Appendix C.
5.3. Dummy Variable Interpretation
Before examining the results of the Granger causality tests and impulse response
functions, the results of the dummy variable coefficients will be presented below. The full VAR outputs for both models are given in Appendix D. For the first VAR model, none of the dummy variable coefficients are statistically significant. This can be seen by observing the t-statistic in square brackets under the coefficients in the output tables. Since the t-statistics are less than 2 in absolute value, the null hypothesis claiming that the coefficient is statistically zero is accepted.
For the second VAR model, the only statistically significant dummy coefficient is the one in the equation for public external debt. With a t-statistic greater than 2 2.0885 , the null
hypothesis claiming that the coefficient is statistically equal to zero is rejected. According to the regression output in appendix D, the value of this coefficient is 322.6145. Since public external debt is differenced in the VAR model, the interpretation of this coefficient is as follows:
Lebanon’s public external debt, on average, was increasing by 322.6145 million US$, after the start of the Syrian crisis, than it was pre-crisis. This supports the hypothesis that the Syrian crisis, and in turn the influx of refugees, has put a sizeable strain on public services, increasing the government’s need of financing. Due to this increased need of financing, the Lebanese government has increased its borrowing from abroad to finance the refugee’s needs. Since Lebanon’s public external debt has increased after the start of the crisis, depreciation pressure may well have resulted for the Lira. As the government borrows more from abroad, increased demand for foreign currencies results, and in turn applies depreciation pressure on the Lebanese Lira. Furthermore, the increased public external debt also aggravates Lebanon’s current account deficits, heightening the importance of sustaining the peg to the dollar. Thus, according to the
39
above results, one may logically presume that the Syrian crisis has increased the pressure on the Lira by increasing Lebanon’s public external debt.
5.4. Granger Causality Tests
We now perform the Granger causality tests to determine whether the number of Syrian refugees in Lebanon Granger causes any of the economic variables. If the influx of refugees Granger causes any of the economic variables, then past values of Syrian refugees in Lebanon help predict current and future values of the economic variables. The results are presented with the chi-square statistics along with their respective p-values. The null hypothesis states that variable X does not Granger cause variable Y. The focus of these results is to examine whether the influx of refugees Granger cause the economic variables. Thus, any other results will not be analyzed.
5.4.1. VAR model 1
For the first VAR model, the results are found below. Each table presents the results of the Granger causality test for each variable. In other words, for every table there is one
dependent variable, and the results reveal whether the other variables in the model granger cause this dependent variable.
40
According to table 4, the influx of refugees does not Granger cause the level of economic activity in Lebanon, since the p-value of the refugee variable exceeds 0.1 (0.1251). Thus, the null hypothesis is accepted.
The results in table 5 indicate that the influx of Syrian refugees in Lebanon also does not Granger cause the amount of foreign direct investment taking place in Lebanon. This conclusion is supported by observing the p-value of 0.9933 which exceeds 0.1.
Furthermore, according to table 6, there is an absence of Granger causality between the change in the number of refugees and the level of portfolio investment in Lebanon, since the p-value of 0.9388 is also larger than 0.1.
However, the results in table 7 indicate that the influx of refugees in Lebanon Granger causes the accumulation of foreign exchange reserves in Lebanon. This can be seen by observing the p-value of 0.0414, indicating a rejection of the null hypothesis.
In general, the above results indicate that the number of refugees in Lebanon does not Granger cause the capital inflows in Lebanon portrayed by portfolio investment and foreign direct investment. This does not support the hypothesis argued in the previous chapter that the inflow of refugees and the war environment neighboring Lebanon dampened the investment climate in Lebanon, leading to a reduction or reversal of the capital inflows. However, past values of refugees in Lebanon does seem to help predict future values of foreign exchange reserves held by the BDL. This supports the hypothesis that the influx of refugees, and the war environment neighboring Lebanon, are seen as a negative shock to the economy by the BDL. Since the sign of the Granger causality is not known, the above result can be interpreted as follows. As the number of refugees increase in a certain period, the BDL sees the danger of
