University of Amsterdam
MSc in Economics
Monetary policy and Banking
Master’s thesis under the supervision of Dr Ed Westerhout
Ville Strandman
(10827218)
BANKING CRISIS AND THE FISCAL
MULTIPLIERS: EVIDENCE FROM THE
NORDIC CRISIS
ABSTRACT
The most recent economic and financial crisis has increased the economic interest in fiscal policies. Especially, the estimation of fiscal multipliers during periods of recessions and financial turbulence has raised several theoretical as well as quantitative questions about the policy makers’ ability to use fiscal policies to smooth out the real business cycle fluctuations. This thesis employs the empirical methodology introduced by two IMF economists, Oliver Blanchard and Daniel Leigh, who identified larger than normal fiscal multipliers during the current crisis in Europe by using two year output growth forecast errors and fiscal balance projections. This thesis investigates whether it is possible to find a similar pattern in an earlier crisis. Combination of the data from the early 1990’s Nordic crisis and the most recent banking crisis provides results which are in line with Blanchard and Leigh’s conclusions. Hence, the findings of this thesis suggest that the higher than normal fiscal multipliers are not just a phenomenon of the most recent crisis.
CONTENTS
1 INTRODUCTION ... 1
2 THE NORDIC CRISIS ... 3
2.1 DEFINING THE BANKING CRISIS ... 3
2.2 ROAD TO THE CRISIS ... 4
2.3 CRISIS RESOLUTION ... 5
2.4 THE MACROECONOMIC IMPACT OF THE CRISIS ... 8
3 LITERATURE REVIEW AND THEORY... 13
3.1 DEFINING FISCAL MULTIPLIER... 13
3.2 DETERMINING THE SIZE OF THE FISCAL MULTIPLIER ... 14
3.2 A BRIEF LOOK AT THE LITERATURE ... 17
3.3 FISCAL MULTIPLIERS DURING A CRISIS ... 18
3.4 FISCAL MULTIPLIERS AND THE NORDIC CRISIS ... 19
4 EMPIRICAL ANALYSIS ... 21
4.1 BLANCHARD AND LEIGH’S METHODOLOGY (2013) ... 21
4.2 DATA AND METHODOLOGY ... 24
4.3 EMPIRICAL RESULTS ... 29
4.4 NORDIC CRISIS ... 33
4.4.1 FIRST STAGE... 33
4.4.2 SECOND STAGE ... 36
5 CONCLUSION ... 41
REFERENCES... 42
1
1 INTRODUCTION
Since the Great Moderation, economists have emphasized the importance and capability of monetary policies as main tools in moderating the real business-cycle fluctuations. While the monetary policies have been praised, fiscal policies have been seen as ineffective and way too slow for the purpose. However, during the recent global economic and financial crisis, monetary policies were compromised when nominal interest rates hit zero lower bound and central banks were forced to introduce unconventional monetary policies such as quantitative easing. Simultaneously, the world witnessed many countries adopt large fiscal stimulus packages while others started aggressive fiscal consolidation programs. In 2009, in the middle of the deepest recession in modern economic history, the US senate approved the American Housing and Recovery Act, an approximately 831 billion dollar stimulus package which aimed to push the economy back on its track. (Feyrer and Sacerdote, 2011) Later many European countries followed the US´ example. In Japan, the current Prime Minister Shinzo Abe started a new economic program, titled “Abenomics”, which combines aggressive monetary expansion, large and flexible fiscal policies and structural reforms which were supposed to end Japan’s decade long deflation. (Oguro, 2014) A 10.3 trillion yen stimulus package was used for example for disaster prevention, infrastructure reconstruction and private investment boosting and was expected to boost domestic economy by 2 percents and create 600 000 new jobs. (Bloomberg News, 2013) Furthermore, after the sovereign debt crisis hit some of the European Union member countries, many of them were forced to undertake heavy fiscal consolidation measures in order to bring their national debts back to sustainable levels and arguably stimulate their domestic economies by restoring market confidence. (Lane, 2012) Recently, even countries without excessive debt problems, such as Finland and the United Kingdom, have started to plan fiscal consolidation. Followed by dramatic changes in fiscal policies, many countries have ended up finding themselves in unexpected positions. Impacts of fiscal policies have been greatly misestimated, which has encouraged economists to rethink the impacts of fiscal policies, especially in periods of economic turbulence. Also, the realization of the limits of monetary policies has brought fiscal policies back to the centre of economic research and interest. The increased researched on fiscal policies has also resulted in a renaissance of the literature on fiscal multiplier.
Fiscal multiplier usually refers to a change in national output caused by a change in fiscal policy. For example, how much will real GDP change if government increases its spending by 1 euro? Many academic papers have estimated fiscal multipliers by using structural VAR (Vector Autoregression) models while others have estimated exogenous fiscal shocks by using data from large fiscal policy changes such as military build ups. In 2013, two IMF economists, Oliver Blanchard and Daniel Leigh, published a paper “Growth Forecast Errors and Fiscal Multipliers” which uses real GDP forecast errors to estimate whether fiscal consolidation has a larger effect on real output during the economic crisis compared to non-crisis periods. They used data from European countries from the recent economic crisis. This thesis tries to contribute to the existing literature by employing their methodology and combining data from the recent
2
financial crisis with the data from the early 1990’s Nordic banking crisis. Recently, many papers (Ilzetzki et al., 2011; Spilimbergo et al., 2009; Batini et al., 2014 among others) have emphasized the importance of country specific characteristics as well as economic status in fiscal policy transmissions. For example, a closed economy with low public debt may be significantly more sensitive to the economic stimulus than a small open economy with large public debt. Nordic countries are relatively homogenous, small open economies with large and heavy public sectors and they experienced a systemic banking crisis simultaneously. By investigating the effectiveness of fiscal policy during the early 1990´s crisis period in these countries, it is possible to gather important information and deepen the understanding on fiscal policy transmission in the economies.
It turns out, that by combining data from the Nordic and the recent crisis increases the statistical significance of coefficient which suggests that similar patterns can be found in both crises. Especially when only countries which experienced systemic banking crisis were included into the sample, the results were notably evident than with the full dataset. Hence, results presented in this thesis suggest that some characteristics of the economy and especially turbulence in the financial and banking sector, but not necessarily the recession itself, seem to result in higher fiscal multipliers.
The structure of this thesis is as follows: First, the build-up, resolution and macroeconomic impact of the Nordic crisis are discussed in great detail. Then, fiscal multipliers are discussed on a theoretical level focusing especially on reasons why economic crisis is expected to amplify the impacts of fiscal policies. After that, the empirical estimations for fiscal policy impacts are performed and results are analyzed. Finally, main implications are concluded.
3
2 THE NORDIC CRISIS
The Nordic Banking Crisis in early 1990´s has been titled as one of the “Big Five”, the most disastrous banking crises before the recent crisis. (Reinhart and Rogoff, 2008) It was also the first systemic banking crisis in advanced economies since the Great Depression in 1930´s. (Sandal, 2004) Three out of four mainland Nordic countries (Finland, Sweden and Norway) experienced systemic banking crisis while Denmark, despite of heavy losses recorded by the major banks and severe liquidity problems in the system, managed to avoid crisis to evolve systemic. In this section the reasons behind the crisis, macroeconomic consequences as well as the resolution for the crisis are discussed. The focus is mainly in three countries (Finland, Norway and Sweden) which experienced systemic crisis but also the Danish case, and especially the reasons why it avoided the systemic crisis, is discussed briefly.
2.1 DEFINING THE BANKING CRISIS
Defining a banking crisis is particularly problematic because the crisis can take many forms, include an undefined number of entities and the publicity of the crisis may vary across the time. Furthermore, due to the lack of public information on banks’ solvency statuses, it is difficult to define when financial distress turns into a banking crisis or becomes systemic. That is why trying to find a universal definition may have several flaws. In general, banking crises can be divided roughly into two categories: bank insolvencies and systemic banking crisis. Reinhart and Rogoff (2008) define systemic banking crisis as an event where bank runs “lead
to the closure, merging, or takeover by the public sector of one or more financial institutions”. According to
their definition milder financial distress i.e. non-systemic crisis is an event where mergers or government support signal that similar outcomes could happen to other institutions as well even if there are no significant bank runs in the system.
On the other hand, Laeven and Valencia (2012) define financial distress systemic if two conditions are fulfilled. First, there has to be signs that the country´s financial system is in heavy distress combined with significant bank runs and credit losses (5 percent monthly deposit decline). Second, there has to be a significant policy intervention as a response to the bank losses. They define a significant policy intervention as monetary authority´s liquidity support that exceeds 5% of the total deposits and is at least double compared to the previous year. In this thesis, the definition from Laeven and Valencia is used to identify a systemic banking crisis.
4
2.2 ROAD TO THE CRISIS
In all three cases, the basic structure of the crisis build-up was similar; deregulation of the financial system, boom in the credit markets and finally an adverse economic shock which triggered the crisis in the banking sector. The financial deregulation in the 1980´s enabled rapid credit growth in all three countries which turned into a bust due to a mixture of negative economic shocks.
Credit growth was fastest in Finland. Foreign banks were allowed to open subsidiaries in Finland in 1982 which increased competition in the credit market, and in 1984 banks were allowed lend abroad and invest in foreign assets. Furthermore, in 1987 HELIBOR (Helsinki Interbank Offered Rate) was introduced, Bank of Finland began open market operations in bank certificate of deposits in the money market and down payment requirements on housing loans and consumer credit were abolished. While previously in an extensively regulated environment households´ borrowing was subject to close bank relationship and personal contacts, financial liberalization and especially a lift of bank lending and deposit ceiling liberated many households from credit constraints. Also, removal of the average bank lending rate in 1986 heightened competition in the banking sector and encouraged small and medium sized banks to expose themselves to excessive risk taking. For example, Skopbank, the central bank of Finnish savings banks, increased its credit growth by 50 percents in 1987 and maintained high growth rate in 1988 and 1989. Commercial banks overall expanded their lending up to 30 percents in 1988. Loans-to-nominal GDP ratio increased from 55 percents in 1984 to 99 percents in 1990 and households´ saving ratio declined from 5.7 to -1.6 in 1980-1988. Also, improved access to credit had effect on the corporate level. Leverage ratios increased and especially borrowing in foreign currency increased rapidly. Before the crisis more than half of the corporate borrowing was determined in foreign currency. (Drees and Pazarbasioglu, 1998)
Similar patterns can be seen in the Swedish deregulation process even though the credit growth was not as significant compared to Finland. Swedish banks´ deposit interest rate ceiling was abolished already in 1978 and in 1980 foreigners were allowed to hold domestic shares. In 1983 the requirements on banks´ bond holdings to meet the liquidity obligations were overthrown. Foreign banks were allowed to open subsidiaries in Sweden in 1986 and during 1986-1988 the foreign exchange controls were loosened and in 1989 removed completely. In 1990 foreign banks were allowed to open branches on Swedish soil. The rapid financial liberation increased Swedish loan-to-GDP ratio from 41 to 58 percents in 1984-1990 which was a much more moderate pace compared to Finland. Households´ net saving rate dropped from 5.0 to -3.4 in 1980-1987 and simultaneously companies increased their leverage levels. Already in 1980 the debt-to-equity ratio of Swedish companies was on average 5.5 and more than 20 times higher than in the United Kingdom (0.2) or in the US (0.25). A notable point about the ratios is that they increased sharply after the deregulation and peaked just before the crisis. Heightened competition in the banking sector encouraged banks to expand their balance sheets and in 1985-1990 the combined balance sheet of all Swedish banks was increasing on average 15 percents a year. (Drees and Pazarbasioglu, 1998)
5
In Norway regulation was radically stripped already in the 1970’s. In 1972 Norwegian banks’ capital adequacy requirements were decreased from 8 to 6.5 percents. Foreign borrowing was partly liberated in 1980 and in 1984 the supplementary reserve requirements were abolished. Furthermore, in 1987 the primary and secondary reserve requirements were abandoned. In 1989-1991 the remaining foreign exchange regulations were erased. Deregulation had an immediate effect on the Norwegian economy: Bank loans-to-GDP ratio increased from 40 to 68 percents in 1984-1988. Households’ net savings in percentages from disposable income dropped from 5.2 to 2.5 and in corporate sector companies increased their leverage ratios. The increased competition encouraged new entities to enter to the market which led to the share of state owned banks to drop from 40 to 20 percents in the 1980´s. (Drees and Pazarbasouglu, 1998)
Overall, the financial liberalization was accompanied by weak and out-of-date prudential policies and the failure to adjust legislation to rapid financial development. (Haugh et al., 2009) Heightened competition and improved access to credit increased financial institutions’ appetite to risk and led to excessive risk taking. Monetary policy was loose in all three countries and mainly driven by stable foreign exchange rate policy which also further encouraged foreign lending.
2.3 CRISIS RESOLUTION
Before discussing further the methods of crisis resolution and the macroeconomic costs, it should be briefly explained why the financial distress in Denmark did not reach a systemic level. Unlike Norway, Finland and Sweden, Denmark was the only country which managed to avoid a systemic crisis even though its banking sector was severely hit. One reason why the crisis did not reach a systemic level in Denmark was the early improvements in the financial regulation. For example, the removal of tax deductibility of debt finance moderated the leverage growth before the economy overheated. Furthermore, the overall prudential policies together with capital adequate requirements were directly stricter compared to the other three Nordic countries. (Honkapohja, 2009) Even though the financial distress in Denmark began already in 1984 when Kronebanken, the 7th largest bank in Denmark, reported heavy losses the main crisis period is generally dated to years 1987-1993. Between 1985 and 1995 105 small or medium sized banks quit their operations and more than half of them because of financial and liquidity problems. Furthermore, banks’ loan write downs increased from 0.5 percent in 1987 to 2.5 percents in 1992 and the overall cumulative bank losses were 9 percents of total loans in the years 1990-1992. (Abildgren and Thomsen, 2011; Reinhart and Rogoff, 2008) In June 1992 Unibank, the second largest bank in Denmark, applied and received liquidity support from the Danish Central bank after rumors about special measurements against the bank by the Danish Financial Supervisory Authority spread. Situation calmed quickly and a large scale bank run was avoided. (Abildgren and Thomsen, 2011)
Overall, two significant differences can be found when the Danish crisis is compared to the other three Nordic countries. First, small and medium size banks were often merged with healthier and larger banks
6
instead if being overtaken by the Danish government. The crisis was mainly handled within the banking sector and public takeovers were targeted to small and medium sized banks. Second, the pre-crisis capital adequacy requirements ensured that the banks were less likely to suffer from liquidity problems even if they suffered from credit losses. The aggregate banking sector remained relatively strong, liquidity problems were isolated to individual banks and the majority of large banks were able to raise fresh capital from the markets without public support. Hence, the public support required was significantly smaller compared to other Nordic countries and thus the crisis wasn’t systemic.
In the other three Nordic countries, the credit fuelled boom ended in a burst of the credit bubble triggered by a mixture of adverse economic shocks. The banking crisis that followed hit Finland the hardest. In the late 1980´s economic development and easy access to credit caused a significant boom in the real estate market and in private consumption which overheated the Finnish economy. Still in 1989, on the eve of the crisis, the real GDP growth was 5.4 percents. The collapse of the former Soviet Union in 1991 caused a 70 percents drop in the Finnish trade with Russia. The Finnish economy relied strongly on the Soviet trade which had lead to its productivity lagging behind the European standards. Furthermore, the reunification of Germany caused a rapid rise in the interest rates all over Europe which put an upward pressure on the Finnish interest rates and led to a significant increase in the real interest rates as well. Increased domestic wage levels, appreciated currency and the absence of Soviet trade caused a significant decline in Finland´s terms of trade in the goods market. Bank of Finland responded to the increased risk of speculative attack on the Finnish Mark by increasing the nominal interest rate but eventually the currency had to be devalued in 1992.
The economic downturn worked as a trigger to the crisis in the banking sector and the first Finnish bank to struggle was Skopbank, the central bank of Finnish saving banks. Skopbank had been under increased public scrutiny already from 1989 but in September 1991 it ran out of liquidity and was bailed out by Bank of Finland, which later took full control over the bank. After the ad-hoc measures, Finnish government decided to step in and in early 1992 it established the Government Guarantee Fund (GGF) which became the main authority to handle the banking crisis. GGF´s main function was to inject fresh capital into the banking system under an approval control of the Finnish government. In August 1992 the government made a promise to meet all loan obligations of the Finnish banking system in all circumstances and a year later the GGF had to be strengthened with additional capital. Overall, the GGF supported 41 banks during the crisis and the Finnish government took over the three largest banks which together held 31 percent of all deposits in the system. (Caprio and Klingebiel, 1996; Reinhart and Rogoff, 2008) The situation in the Finnish banking sector improved slightly at the end of 1993 but the public measures continued through 1994. Yet, it took until 1997 for banks to turn the course again to significant profits. (Honkapohja, 2009; Sandal, 2004)
In the late 1980´s, the Swedish economy was not as strongly dominated by the Soviet trade as the Finnish one. The absence of strong trade bonds with the Soviet Union had induced the Swedish industry to modernize its machinery to improve productivity. Later these reforms moderated the recession and helped
7
Sweden to recover faster compared to Finland. However, the Europe wide interest rate hike resulting from the unification of Germany caused a currency crisis and the following recession can be seen as a trigger for the banking crisis in Sweden. In autumn 1991 Första Sparbanken, the largest savings bank in Sweden, and Nordbanken reported heavy losses. In the early crisis, the countermeasures were treated from an ad-hoc base with additional capital injections. However, it took until the summer of 1992 for the crisis to become systemic. In 1992, the seven largest banks in Sweden, which together held 90 percents of all assets in the system, reported heavy losses. Situation was worsened by a capital outflow when foreign investors withdrew their investments out of Sweden. On September 24th 1992 Swedish officials made a parliamentary announcement to guarantee the loan obligations of all banks and the Swedish central bank, Riksbanken, started to use its foreign currency reserves for liquidity injections to the banking system. In 1993, Bankstödsnämnden (BSN), equal to the Finnish GGF, was established to provide public support to the weakening banking system. BSN provided public support to all banks which were willing to apply for it, as was done by all of the large banks, excluding Svenska Handelsbanken. The capital and liquidity injections improved the situation and the confidence in the banking sector started to restore quickly. However, certain supportive measures remained for a long time after the crisis and for example the guarantee blanket for banks’ loan obligations was kept until July 1996. (Honkapohja, 2009; Sandal, 2004)
The Norwegian economy experienced an economic recession earlier than Finland and Sweden but it took several years for the crisis to hit the banking sector and to evolve to a systemic level. In 1986, a rapid drop in the global oil price dragged the Norwegian economy into a severe recession. In 1985, the global oil price for a barrel was still over 27 dollar but dropped sharply below 12 dollars in the third quarter of 1986. (OECD World Economic Outlook June 1986) The price shock caused a dramatic drop in Norway´s oil revenue and turned the country´s current account to GDP ratio from 4.8 percents surplus in 1985 to 6.2 percents deficit in 1986 which led to speculative attacks on the Norwegian Krona. As a response, the Norwegian Krona had to be devaluated by 6 percents in 1986. (Vale, 2004) In the autumn of 1988, Sunnmörstbanken, a medium sized commercial bank, announced losses larger than 25 percent of its equity capital and in 1989 it was followed by Norion bank, which reported it had lost all of its capital. Later Norion was put under full public control. Between 1989 and 1990, 11 local and regional savings banks received support from The Savings Banks Guarantee Fund (SBGF) and they were merged with larger and still solvent banks. The situation was handled mainly within the banking sector until 1991 when SBGF’s funds were running low and the losses in the banking sector were still increasing. At this point, the Norwegian government stepped in and established a national crisis management authority, the Government Bank Insurance Fund (GBIF), which initial responsibility was to support two private sector fund guarantee entities so that they could carry out their purpose. During 1991, the further worsened situation in the banking sector led to more direct capital injections to the banks. In the autumn of 1991, the second and the third largest commercial banks, Christiania Bank and Focus Bank respectively, required fresh capital while the largest bank, Den Norsk Banks, recorded significant losses. Together these banks held 54 percents of all assets in the banking system which meant that
8
the crisis had reached a systemic level. The government responded by increasing the capital of the GBIF and by establishing the Government Bank Investment Fund (GBF) which was supposed to provide liquidity to the banks based on their commercial evaluation. This was to address the increasingly illiquid wholesale banking market which was depressed by losses of the three largest banks. In the spring of 1992, all three banks were nationalized and the value of their old shares was written down. In 1993 the situation in the Norwegian banking sector started to improve quickly. In the post-crisis period, the Norwegian government gradually and successfully privatized the nationalized banks and managed to net benefit from the whole process. In other words, despite of the publically funded capital injections, the Norwegian government managed to increase enough funds from the privatization to cover all of its fiscal costs and profit from the nationalization. (Honkapohja, 2009; Sandal, 2004)
2.4 THE MACROECONOMIC IMPACT OF THE CRISIS
The impact of the banking crisis alone on the Nordic economy is difficult to evaluate because the crisis was triggered by a mixture of negative economic shocks. In general, the economic downturn is expected to be longer lasting and output loss to be higher during a banking crisis than during a normal recession. (Haugh et al.2009) Bernanke and Gertler (1995) highlight the important role of banks in the credit markets and suggest that a disruption in bank lending may lead especially the small and medium sized companies to be if not completely excluded at least significantly credit constrained. This may have a negative effect on real activity in the economy and amplify the negative impact of the economic shock. Hence, it should be noted that the economic impacts described below are most likely caused by a combination of business cycle fluctuation, economic distress and a failure in the banking sector. Furthermore, since the definition of a banking crisis is broad, it is difficult to determine the exact dates of the crisis. In this thesis, years 1990-1993 are considered as the main crisis period because the most of bank losses were recorded in these years. According to Laeven and Valencia (2008) banking crisis became systemic in Sweden, Finland and Norway in 1991, but the financial distress and problems in the banking sector began for example in Norway already in late 1980’s and in Finland and Sweden in 1990. During the crisis period, at worst 13 percents of the bank loans in Finland and Sweden were compromised while in Norway the value was 16.4 percents. Loan losses compared to GDP during the period were 8.0, 6.4 and 4 percents in Finland, Sweden and Norway respectively. (Laeven and Valencia, 2008)
Graph 1 shows the annual real GDP growth development in all Nordic countries during 1980-2000. In Finland the impact on real GDP was the most significant and the output loss was almost eight times larger than during normal recessions. (Haugh et. al, 2009) For Finland, the cumulative output loss during the whole crisis is estimated to be 59.1 percents and the lowest annual real GDP growth -6.2 percents was recorded in 1991. Already in 1990 there was zero economic growth and the economy kept shrinking until 1993. The recovery on the other hand was remarkably fast and in 1994 the real growth was already 3.9 percents. The output drop was second highest in Sweden and the cumulative output loss was 30.6 percents from GDP. At
9
worst the Swedish economy was declining at an annual rate of -3.8 percents. The Finnish industry began an aggressive catch-up process after the Soviet trade collapsed and already in 1993 the Finnish economy had overtaken the Swedish one in the recovery process. As explained above, Norway experienced an economic crisis earlier than Sweden and Finland but the recession worked as a trigger for the banking crisis. In the Norwegian financial sector small and medium sized banks were severely distressed already during the recession which could have resulted in the sluggish recovery. Yet, the Norwegian economy did not suffer from any significant output losses during the crisis. In 1993, the annual real growth pace dropped from 3.4 to 2.3 but the drop could have reflected the decline in the global demand which followed the higher global interest rates. During the economic recession the Norwegian economy’s decline was at its worst in 1988 when the growth rate was -0.5 % and the lowest growth rate during the period of systemic banking crisis was 2.8 percents.
As graph 1 illustrates, the economic recession in Denmark took place at the same time as in Norway. Danish economy experienced eight years of sluggish economic growth in 1985-1993 and in those years real GDP growth never exceeded 2 percents. The recession in the second half of the 1980’s was strongly driven by normal real-business cycle fluctuation and the slow recovery could be partly attributed to worsened international competition position of Denmark in the mid 1980’s due to increased labor costs in the early 1980’s. Graph 1
-5
0
5
G
rowt
h
1980
1985
1990
1995
2000
Year
Finland
Sweden
Norway
Denmark
Source: OECD World Economic OutlookReal GDP Growth Rate
Nordic Countries 1980-2000
Gro
w
h R
at
e
(%
)
10
Graph 2 illustrates the unemployment rate development during the Nordic crisis. The impact of crisis on employment was largest in Finland and there was a steep increase in unemployment beginning of 1990 until 1994-1995 while the recovery was moderate from 1995 onwards as the graph illustrates. Similar trends can be seen in Swedish unemployment development even though it took until 1997 for unemployment to decline significantly. The slow recovery of the Norwegian economy is well illustrated in the unemployment figure. The unemployment was steadily increasing from 1988 and peaked in 1993. However, it should be noted that the unemployment rate at its worst barely exceeded five percents which can be explained by the low output response of the Norwegian economy to the turbulence in the banking sector. In Denmark on the other hand, the unemployment was steadily increasing from 1986 onwards and turned around as late as 1993. Increasing unemployment rates can be explained by slow recovery and growth rates in early 1990’s.
Graph 2
Graph 3 illustrates the current account balances in US dollars for Nordic countries from 1980-2000. It seems that almost every country had a current account deficit at the second half of the 1980’s which decreased further until the bottom was hit in the early 1990’s. As discussed above, Finland was hit hard by the collapse of Soviet trade in the early 1990’s. The significant drop in exports caused a large decline in the current account and a major current account deficit. Similar development, though not equal in magnitude, can be seen in the Swedish data. In both countries the deficit was turned into a surplus in 1993 which signals a
0
5
10
15
20
Un
em
pl
oy
m
en
t
1980
1985
1990
1995
2000
Year
Finland
Sweden
Norway
Denmark
Source: OECD World Economic OutlookUnemployment Rate
Nordic Countries 1980-2000
U
nem
pl
oy
m
en
t R
ate
(%
)
11
strong performance and importance of the export sector in the economic recovery. The impact of the Nordic crisis was after all restricted to a small number of countries and the decision to devaluate domestic currencies in 1992 launched remarkable growth in exports especially in Sweden and Finland. (Jonung, 2009) In Norway, the current account balance was turned from 4.8 percent surplus in 1985 into 6.2 percent deficit in 1986 by the decline in global oil price. Later this forced the Norwegian central bank to devaluate its currency by 6 percents which improved the competitiveness of the economy. It should be noted however, that apart from a drop in oil revenue, Norway´s external balance remained highly positive through the crisis period. (Vale, 2004) In Denmark, as in Norway the most significant drop happened in 1986. An increase in the current account deficit was partly due to highly increased real wages in the overheated Danish economy in the late 1980’s during which the wages increased 10 percents annually. Even though the drop was relatively sharp the Danish export sector recovered quickly and was further boosted by the economic boom in Germany.
Graph 3
In general, the impact of the banking crisis in the different countries varied both in time as well as in magnitude. All Nordic countries can be characterized as open small economies with relatively heavy public sectors. Finland took no doubt the heaviest hit, which can be partly explained by the close trade bonds with the former Soviet Union. The Swedish economy experienced a similar rapid drop in real GDP as Finland
-10
00
0
0
10
000
20
000
30
000
C
u
rre
n
t
Acco
u
n
t Ba
la
n
ce
1980
1985
1990
1995
2000
Year
Finland
Sweden
Norway
Denmark
Source: OECD World Economic OutlookCurrent Account Balance in Millions of US$
12
although not as severe. In these countries the recovery took place remarkable fast due to their strong export sectors and the strong economic situation in the rest of Europe. The Norwegian economy had the recession before Sweden and Finland but ended up suffering from a prolonged recovery partly because of problems in the banking sector. Arguably its proper recovery began after the banking crisis. Denmark suffered from the spillover effect from the rest of the Nordic region. Furthermore, the turbulence in the financial sector, though not at systemic level, may have slowed down its recovery as well. Especially in Finland and Sweden the recovery can be attributed to structural reforms in the labor markets and in industry, the increased competitiveness of domestic industry and strong performance of the export sector.
13
3 LITERATURE REVIEW AND THEORY
In this section, fiscal multipliers are discussed in detail. First, a definition for fiscal multiplier is provided followed by a brief theoretical and literature insight which focuses especially on factors which may effect the size of fiscal multipliers on a general level. Next, the arguments for higher multipliers during a recession are discussed and the reasons why it can be expected that fiscal multipliers were amplified during the Nordic crisis are explained.
3.1 DEFINING FISCAL MULTIPLIER
A very basic definition for fiscal multiplier is a marginal change in the GDP or other measure of output caused by an exogenous marginal change in the government’s discretional fiscal variable. (Ilzetzki et al. 2011) For example, if the Finnish government increases its spending by 1 euro and it causes the GDP to increase by 1.5 euros, the fiscal multiplier for Finland equals 1.5. Government spending is often used as an example of fiscal shock but in reality a change in the fiscal policy could include changes in tax rates or combination of the two. (Spilimbergo, 2009) Spending multiplier refers to an output response caused by a change in a government’s spending or investment. Tax multiplier means output change caused by a discretionary change in a government’s revenue and overall multiplier describes the output response caused by an unspecified fiscal shock.
There are several different ways to measure fiscal multipliers but the two most used are impact multiplier and cumulative multiplier. Impact multiplier measures an economy’s output response within the period when a fiscal policy change occurs. Impact multiplier can be expressed more formally as:
𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 𝑀𝑀𝑀𝑀𝑀𝑀𝐼𝐼𝑀𝑀𝐼𝐼𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 =∆𝐺𝐺∆𝑌𝑌𝑡𝑡
𝑡𝑡 (1)
where ∆𝑌𝑌𝑡𝑡 indicates a change in the output at period t and ∆𝐺𝐺𝑡𝑡 means a change in the fiscal variable. Cumulative multiplier is used when the fiscal policy change is wanted to measure at a longer time period. It measures the cumulative effect of a fiscal shock over time period N and is a widely used measure of the effectiveness fiscal policy. Cumulative multiplier can be expressed formally as:
𝐶𝐶𝑀𝑀𝐼𝐼𝑀𝑀𝑀𝑀𝐼𝐼𝐼𝐼𝑀𝑀𝐶𝐶𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀𝐼𝐼𝑀𝑀𝐼𝐼𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 =∑∑𝑁𝑁𝑗𝑗=0∆𝑌𝑌𝑡𝑡+𝑗𝑗
∆𝐺𝐺𝑡𝑡+𝑗𝑗
𝑁𝑁 𝑗𝑗=0
(2)
where ∑𝑁𝑁𝑗𝑗=0∆𝑌𝑌𝑡𝑡+𝑗𝑗 indicates a cumulative change in the output over a period N and ∑𝑁𝑁𝑗𝑗=0∆𝐺𝐺𝑡𝑡+𝑗𝑗measures a cumulative change in the fiscal variable over a period N. In this paper it is referred to cumulative multiplier unless otherwise noted.
14
3.2 DETERMINING THE SIZE OF THE FISCAL MULTIPLIER
The economic literature which tries to estimate the size of fiscal multiplier is enormous and has divided the community of economists for decades. Despite of an absence of a universal value for the multiplier, the general status quo states that the size of the multiplier is determined by several factors and economic characteristics. These can be further divided into two sub-categories, permanent and temporary factors, from which the latter is more relevant in the context of this thesis. (Batini et al., 2014) Permanent country specific factors refer to structural characteristics which determine how fiscal policies are implemented and via which transmission channels they impact the real economy. These include foreign exchange rate regime, labor market flexibility, the level of sovereign debt, automatic stabilizers and economy´s openness to trade. Temporary factors refer to non-structural circumstances which may increase or decrease the multiplier. These include the state of the real-business cycle, zero-lower bound of nominal interest rate and the condition of financial system. First, the permanent factors are discussed before focusing on the temporary factors.
In an overly simplified version of Keynesian economics, where output is assumed to be a sum of private consumption (C), investment (I), government spending (G) and foreign trade balance (NX), the fiscal multiplier could theoretically equal one. In such a case, the fiscal impact is restricted to the “G” component which has a direct linear effect on output (Y) i.e. Y=C+I+G+NX. However, such a generalization usually leads to over simplification and rarely describes an economy accurately. Even a simple structural modification to the model is likely to result in the multiplier deviating from unity. For example, if the government finances its spending by raising new debt, the increased demand for credit drives up the interest rate. The inclined interest rates may encourage private savings and lower private consumption. Furthermore, the higher interest rate may discourage private investment which has a negative effect on output. In such cases, it is said that the government crowds out private investment which results in the fiscal multiplier being less than one.
If fiscal policy changes have effect on households’ real income in the economy, large automatic stabilizers could smooth the effect via income transfers and taxation. Automatic stabilizers are components in the fiscal policy which mitigate the output fluctuation without intentionally planned policy changes by the government. Dolls et al. (2010) suggest that if a fiscal policy change causes a shock to the households’ gross income then two factors determine the impact on disposable income; the tax and transfer system in the economy and the relation of disposable income and the current demand for goods and services. The first factor indicates that if the income tax rate in an economy is 40 percents, then the shock to households’ disposable income is mitigated to just 60 percents. On the other hand, the second factor is closely related to the permanent income hypothesis which suggests that the temporary changes in income may have a small effect on household’s consumption pattern if the change has only a small effect on permanent (or lifetime) income. Hence, households can use savings or credit to smooth their consumption through their lifetime. If
15
for example an increase in the transfer payments is considered transitory and the permanent income hypothesis holds, households’ demand remains unchanged, (or changes only a little) and the effect on national output is close to zero. However, if households’ demand is closely linked to their current income, for example because of credit constraints, an increase in disposable income will have a direct impact on demand, thus the fiscal policy has larger effect on output. Dolls et al. estimated that within the European Union the effect of the automatic stabilizers is 38 percents from the actual size of the shock while in the US the effect is only 32 percents. In other words, in Europe, 38 percents (32 percents respectively in the US) of the income shock is absorbed by automatic stabilizers which decreases the effectiveness of fiscal policy changes on output.
Another permanent factor affecting the fiscal multiplier is if an economy’s exchange rate is predetermined i.e. if the country belongs to a fixed exchange rate regime or a currency union and therefore its ability to use fiscal policy effectively increases. Intuition follows from the standard Mundell-Fleming model; the expansionary fiscal policy causes a demand shock which shifts the IS-curve to the right and increases the interest rate. The higher interest rate attracts foreign investors and causes capital inflow which increases pressure to appreciate the domestic currency. The appreciated currency decreases demand for domestic exports which crowd outs the expansionary effect on output. However, if the exchange rate is fixed, the monetary authority fully accommodates to the upward pressure on the domestic interest rate by expanding the money supply which diminishes the appreciation pressure. Hence, under a fixed exchange rate regime the net exports do not cancel out the fiscal policy´s impact on output, and thus the domestic demand increases and the fiscal multiplier exceeds unity. (Born et al., 2013) Coenen et al. (2012) tested seven structural DSGE models used by institutions involved in policy making to analyse the effectiveness of fiscal policy under different monetary regimes. They found that fiscal policy became significantly more effective if monetary policy was accommodated by fixing the interest rate to a lower level for even a short period of time. The monetary accommodation supported efficiency of fiscal policy both in temporary short-term expansions as well as in more long-run policy changes. Furthermore, Ilzetzki et al. (2011) estimated the impact multiplier for countries which belonged to a fixed exchange regime to be 0.09 and the long-run cumulative multiplier to be 1.5. Interestingly they found that under flexible exchange rate regime both impact and cumulative multiplier were negative. Hence, their findings suggest that exchange rate regime is one of the fundamental factors determining the effectiveness of the fiscal policy changes on output.
The flexibility of labour markets and the strength of labour unions can make the fiscal multiplier larger. The rigidness of the labour market may cause wage stickiness which increases an economy’s sensitivity to fiscal shocks. If unions have bargained an above equilibrium wage level, involuntary unemployment occurs and the sensitivity of the real marginal cost of labour to aggregate demand of labour decreases. Furthermore, downward sticky wages makes unemployment more sensitive to negative technology shock. Higher unemployment itself may increase the size of the fiscal multiplier if it for example makes households more credit constraint. (Coenen et al., 2012)
16
The level of sovereign debt may also impact the fiscal multiplier. Ilzetzki et al. (2011) estimated that fiscal policy becomes decreasingly effective when a country´s debt-to-GDP level increases and they argued that it could even become counterproductive if the ratio exceeded 60 percents. They estimated impact and cumulative multipliers for countries during periods of high debt levels. To the sample they included countries in which public debt-to-GDP exceeded 60 percents in three consecutive years. They found that both, but especially the cumulative multiplier, were systematically lower compared to periods of below 60 percents debt levels and in some cases they even found evidence of negative multipliers. They suggested that the reason for lower multipliers during the periods of excessive debt was consumers´ expectations of future fiscal policy changes. If for example a highly indebted government increased spending today, consumers’ would expect tax increases in the future and increase their precautionary savings. In such a case, the impact of increased spending would be mitigated by increased savings.
The effectiveness of fiscal policy may also be determined by country´s openness to trade. For example, if a country has low propensity to import goods, part of the fiscal shock is erased by lower net exports and therefore the multiplier is higher. Ilzetzki et al. (2011) estimated that the impact multiplier varies from -0.28 to 0.02 between open and closed economies respectively and the cumulative multiplier varies from -0.75 (open) to 1.29 (closed). The negative multiplier for open economies can be explained partly by the higher price elasticity of imports than exports. Barrell et al. (2012) suggested that in such a case, assuming sticky domestic prices, fiscal shock could turn the multiplier to the opposite sign, i.e. from positive to negative. To summarize, the circumstances which increase the size of fiscal multiplier are a membership of fixed exchange rate regime or currency union, inflexible labour markets and powerful labour unions and economy´s isolation from the international trade. The size of the fiscal multiplier can be decreased by floating exchange rate, high public debt level and high volume of international trade.
The discussion above focuses mainly on factors which may cause the fiscal multiplier to deviate from unity. In some cases, however, empirical evidence from negative multipliers was found, in which case for example an increase in government spending would lower national output. There are some reasons for why fiscal policy would be counterproductive. Perrotti (1999) suggests that the non-linear properties of tax rate change distortions could cause fiscal policy to become counterproductive. For example, if the tax distortion to households’ income increased exponentially with tax rate, households could respond to the government’s increased spending today and potential tax increases in the future by cutting consumption more than proportionally today. This could be the case especially if the sovereign debt level was high. On the other hand, if the government undertakes credible fiscal consolidation measures which lowers the public borrowing premium, the decline in the real interest rate could encourage private consumption and crowd-in private investments. Such policy measure is called “expansionary fiscal consolidation”. Empirical evidence from such economic properties has been found for example by Giavazzi and Pagano (1990) who investigated the fiscal consolidation in Ireland and Denmark.
17
3.2 A BRIEF LOOK AT THE LITERATURE
The size of the fiscal multiplier is an ongoing debate in the economic literature and is still far from being settled. In the case of non-crisis period even the sign of the multiplier has caused some debate. Blanchard and Perrotti (2002) estimated that in general multipliers were closer to 1 and noted that even though private consumption increase after a government spending shock, private investment is likely to be crowded out. Perrotti (2005) estimated that the multiplier would rarely exceed 1 and he could not find support for the claim that tax based fiscal shocks would have any larger or faster impact on real economy than spending based. On the other hand, Alesina et al. (2012) estimated that especially tax based fiscal consolidation could have a severe impact on investors’ confidence and cause a negative effect on real economy. Furthermore, they estimated that the spending based consolidation would lead to more rapid recovery.
During the recent crisis, many countries faced increasing public debt levels which together with a few countries ending up to the verge of default triggered the European wide fiscal consolidation wave. Especially countries with an excessive debt problem undertook radical consolidation measures. The sluggish recovery in countries with large consolidation programs has encouraged increasing volume of research estimating the fiscal multipliers in recessions.
Auerbach and Gorodnichenko (2012) found that fiscal multipliers were significantly more positive during recessions and argued that governments could benefit from using fiscal policy to stabilize cyclical fluctuation. Furthermore, they estimated that the negative side effects of fiscal expansion, such as high inflation, are less likely when the economy is in recession. In favour of active fiscal policy were also Baum et al. (2012) They used data from G7 countries (excluding Italy) and estimated that fiscal multipliers were on average higher in economic downturns (around 2 when the output gap was negative) but they emphasised the variation between economies. They also noted that the results were more consistent for the spending side and that the conclusion for the revenue multiplier was much more ambiguous. Canzoneri et al. (2012) suggested that the fiscal multiplier, indeed, could be counter-cyclical if financially constrained people’s access to liquidity was improved. In other words, they found that the size of the multiplier was determined by the state of the economy and they estimated higher (or more positive) multipliers during recessions. They further argued that the government’s ability to effect on the financially constrained agents in the economy increased the larger the output gap was. They concluded that the recession period multiplier could easily exceed 2 while in booms it was likely to remain below 1. Many economists, de Cos and Moral-Benito (2013) and Owyang et al. (2013) among others have reached a similar conclusion in their research.
18
3.3 FISCAL MULTIPLIERS DURING A CRISIS
Since this thesis tries to investigate whether economic turmoil and banking crises amplify the fiscal multiplier (i.e. makes them more positive), it is necessary to explain the reasons why such an effect could be expected. In the economic literature, at least three explanations for higher multipliers during a crisis can be identified: zero lower boundary of nominal interest rate, poorly functioning financial system and general slack in the economy. Even though it is easy to see these hypotheses to effect on economic dynamics simultaneously they are discussed separately for the sake of clarity. It should be kept in mind however that crises tend to have individual characteristics, which is why these channels can work individually or together. According to classical Keynesian economics an economy becomes increasingly sensitive to fiscal shocks when there is slack in the economy. Intuitively, when the output is below its flexible price equilibrium the fiscal policy is less likely to crowd out private investment by putting an upward pressure on real interest rates. Auerbach and Gorodnichenko (2010) used data from 1947 to 2008 and estimated that during non-recession times the fiscal multiplier was around 0-0.5 while during non-recessions the long-run multiplier could have been almost 2.25 after 20 quarters. In the short-run they estimated the fiscal multiplier to be around 1-1.5 in recessions. On the other hand, Rendahl (2012) took a different approach to economic slack and argued that during recessions unemployment used to peak. He argued that if consumers are forward looking and the unemployment is high and persistent combined with the low interest rate, the economy may drift to a vicious circle where consumers cut spending which leads to increasing unemployment and further decreases demand. If the government steps in with a fiscal stimulus package, it could turn the vicious circle around and make the growth self full-filling. Rendahl estimates that in a severe recession with unemployment exceeding three percentages from the equilibrium level, fiscal multipliers could be around 1.5. However, he notes that the depth of the recession and the size and composition of the fiscal shock may cause some non-linear properties to output deviation. In other words, this would suggest that the deeper the recession is, the larger the multiplier would be.
In mainstream macroeconomics it has been assumed that during normal times an increase in government spending raises inflation expectations for which central bank responses by following Taylor rule and increasing the nominal interest rate. Woodford (2010) suggests that especially if the turbulence in the financial markets has caused interruptions to the financial intermediation and for example recession has pushed output under an efficient level, monetary policy would be unable to further accommodate the nominal interest rate. In such circumstances it is possible that for example fiscal stimulus would increase output without increasing inflationary expectations; hence keeping the monetary policy unchanged. If the interest rate is expected to remain at the zero lower bound (which does not necessarily have to be 0) for a prolonged time, the government’s spending multiplier could exceed 1. Furthermore, it is possible that at zero lower bound, deflation expectations cause an upward pressure on the real interest rate which encourages savings and hence dampens output. (Christiano et al., 2010) Eggertson (2011) suggests that under such
19
circumstances the government multiplier could be extraordinary large if the fiscal stimulus was aimed to the demand side of the economy. He argues that when an economy is at the zero lower bound, it becomes strongly demand driven and he suggests that fiscal policy should be targeted to expand the demand. Intuitively, at the zero lower bound there are not enough buyers which is why increasing production would increase the deflation expectations and cause an incline in the real interest rates. Targeting fiscal policy to boost demand increases overall spending in the economy.
Canzoneri et al. (2012) suggest that counter-cyclical financial frictions may increase economy´s sensitivity to fiscal policy changes. They argue that an adverse economic shock triggers the financial accelerator mechanism and increases credit constraints through the economy. When output increases due to fiscal expansion, the financial frictions mitigate which encourages private consumption and investment. Higher private consumption is assumed to have the same effect as public spending and it further increases output. Increased output level decreases households’ financial constraints and has an increasingly positive effect on the real economy. In other words, the financial accelerator mechanism can be used to make fiscal policy more effective and higher fiscal multipliers are partly due to this effect. On the other hand, Eggertson and Krugman (2012) estimated the New Keynesian model to investigate the effect of a rapid change in the financial environment. They suggested that a rapid and adverse change could lower the financial sectors estimations for household´s credit limits and push them over the debt overhang cliff. Under such circumstances, part of the households become highly credit constrained which puts downward pressure on aggregate demand. Furthermore, the consumption pattern of the credit constrained households becomes more sensitive to current income changes which amplify the effect of fiscal policy changes. Overall, credit constraints, potentially tightened by credit crises, may bring the economy to a liquidity trap which increases households and companies´ marginal propensity to consume and invest. This could make economy increasingly sensitive to fiscal policy changes.
3.4 FISCAL MULTIPLIERS AND THE NORDIC CRISIS
Based on what is known about the economic environment in the Nordic countries during the Nordic crisis and from the discussion above, a certain hypothesis can be made regarding the fiscal multipliers in the Nordic countries in the early 1990’s. If the permanent characteristics are considered, Nordic countries in the early 1990´s can be described as small open economies. Large and heavy public sectors suggest that automatic stabilizers played an important role in income fluctuation via transfer payments and taxes. Furthermore, the countries were not highly in debt and for example in Norway, Finland and Sweden the debt to GDP ratios in 1990 were all below 60 percents. (OECD World Economic Outlook 1990 June) However, even though the countries did not have necessarily strictly fixed exchange rates, the monetary policy was strongly driven by stable exchange rate policy which limited the monetary maneuverability.
20
The above named characteristics were known before the crisis and it is unlikely that they would have surprised forecasters, but how about the temporary and crisis specific factors? When the time period from 1990 to 1993 is considered few reasons which could have led to higher multipliers are highlighted. First, especially in Sweden and Finland, there was slack in the economy. The output drop was significant and the unemployment increased considerably. Also in Norway, the growth had been sluggish since the economic recession in the late 1980´s which suggests that the recovery process was still ongoing. On the other hand, all four countries had severe imbalances in the financial sectors. Banks suffered from significant losses which had a negative impact on credit growth. Even in Denmark, though not in the same scale as in other three countries, the banking sector experienced significant losses and in addition the possibility of spillover effects from its neighbor countries could have amplified the problems in the financial markets. For example, in graph 2 (see section 2) the drop in the real GDP growth in 1991-1992 could have been a collateral impulse from the rest of the Nordic region. Finally, it should be noted that the interest rates in the whole region were nowhere near to zero lower bound. Actually, the development was contrary to what could have been expected, because central banks, especially in Finland and in Sweden, tried to defend the domestic currencies against speculative attacks which led to an increase in the nominal interest rates. Also the real interest rate increased significantly and for example in Finland it peaked in 1993 when it hit 13 percents.
21
4 EMPIRICAL ANALYSIS
In this section, econometrical analysis is used to investigate whether empirical evidence can be found to support the hypothesis that the fiscal multipliers were larger during the Nordic crisis. More specifically, the empirical analysis investigates whether the fiscal policy overall was more effective during the crisis period. As mentioned in the literature review, some previous research has found evidence which suggests that tax changes would lead to higher fiscal multipliers compared to spending based changes (for example Alesina and Ardagna, 2010). However, because the analysis is aimed to identify larger than normal multipliers it does not matter what the actual size of the revenue and spending multipliers were in the pre-crisis period. Furthermore, since the empirical section of this thesis does not make difference between the two policy tools, the results are restricted to the overall multiplier. The reason for focusing on the overall multiplier is partly due to data availability and the chosen methodology. The methodology used is based on a paper from Blanchard and Leigh (2013) which will be discussed in the first part of this section. Next, the methodology and the data used will be introduced and in the final part the results are analyzed.
4.1 BLANCHARD AND LEIGH’S METHODOLOGY (2013)
The empirical work on fiscal multipliers has raised a significant amount of debate about the suitable methodology for the purpose. A common empirical obstacle in the fiscal multiplier focused research is to isolate the exogenous fiscal shock from other fiscal noise caused for example by automatic stabilizers. Furthermore, the direction of causality is often difficult to distinguish, i.e. is it expansionary fiscal policy that increases the output or the increased output that causes an increase in public spending. This is why economists have used various econometrical methods to isolate the fiscal multipliers including for example Vector Autoregression (VAR) and Structural Vector Autoregression (SVAR) models. In this thesis the empirical section follows closely the methodology introduced in the IMF working paper “Growth Forecast
Errors and the Fiscal Multipliers” by Blanchard and Leigh (2013). Rather than estimating the fiscal
multiplier directly, the underlying idea is to estimate whether economic turbulence amplifies the effect of fiscal policy changes. If that is the case and assuming the forecasters did not expect higher multipliers or misestimated them, the amplified impact of policy changes on real economy could be spotted as a deviation of the real GDP growth projections from the realized growth. Furthermore, if economic turbulence amplifies fiscal multipliers it should be possible to identify a systemic relationship between fiscal policy changes and real GDP forecast errors. The methodology is strongly based on the assumption that the forecasters used the correct and accurate model on forecasting which puts right weight on fiscal policy changes during the crisis periods. If the assumption holds, then under rational expectations, the expected coefficient for the non-crisis period should be close to zero.
To test their hypothesis Blanchard and Leigh used standard Ordinary Least Squares (OLS) estimates to regress the two years cumulative real GDP growth forecast error on forecasts of fiscal policy change
22
projected at the beginning of the two year interval, i.e. at time t. By using the two year intervals they take into account the possible lagged effects of fiscal policy change and the slow adjustment of real economy. The baseline regression can be expressed as:
𝐹𝐹𝐹𝐹𝑀𝑀𝑀𝑀𝐼𝐼𝐼𝐼𝐹𝐹𝐼𝐼 𝑀𝑀𝑀𝑀𝑀𝑀𝐹𝐹𝑀𝑀 𝐹𝐹𝑜𝑜 ∆𝑌𝑌𝑖𝑖,𝑡𝑡;𝑡𝑡+1= 𝛼𝛼 + 𝛽𝛽𝐹𝐹𝐹𝐹𝑀𝑀𝑀𝑀𝐼𝐼𝐼𝐼𝐹𝐹𝐼𝐼 𝐹𝐹𝑜𝑜 ∆𝐹𝐹𝑖𝑖,𝑡𝑡;𝑡𝑡+1|𝑡𝑡+ 𝜀𝜀𝑖𝑖,𝑡𝑡;𝑡𝑡+1 (3)
in which Y denotes GDP and F denotes fiscal balance. The left hand side of the expression represents the cumulative two years real GDP growth forecast error for country i. The two years cumulative real GDP,
∆𝑌𝑌𝑖𝑖,𝑡𝑡;𝑡𝑡+1, is measured as (𝑌𝑌𝑖𝑖,𝑡𝑡+1⁄𝑌𝑌𝑖𝑖,𝑡𝑡−1− 1) where 𝑌𝑌𝑖𝑖,𝑡𝑡+1 indicates the real GDP value of country i in the
period t+1. Hence, the cumulative change takes place in period t and t+1. For example, the two years cumulative growth projection for year 2010, t, includes the projected change in the value of real GDP from end of 2009 to end of 2011, i.e. t-1 and t+1 respectively. The associated forecast error is measured
as ∆𝑌𝑌𝑖𝑖,𝑡𝑡;𝑡𝑡+1− 𝑜𝑜(∆𝑌𝑌𝑖𝑖,𝑡𝑡;𝑡𝑡+1|Ωt) where the forecast of the two years cumulative real GDP growth is conditional
to the information set Ω available at the beginning of year t. In other words, the projections for both fiscal balance and real GDP growth are for t and t+1 and projections are made in the beginning of year t.
The forecast of 𝛥𝛥𝐹𝐹𝑖𝑖,𝑡𝑡;𝑡𝑡1 indicates a projected change in government fiscal balance. The projected change is measured by estimating how much the fiscal balance changes in the two year period, i.e. if the projection has been made in 2010, how much the fiscal balance is expected to change from 2009 to 2011. Formally this can be expressed as f(𝐹𝐹𝑡𝑡+1;𝑖𝑖− 𝐹𝐹𝑡𝑡−1;𝑖𝑖|Ωt) where the ‘f’ stands for forecast, 𝐹𝐹𝑡𝑡+1;𝑖𝑖− 𝐹𝐹𝑡𝑡−1;𝑖𝑖 indicates the change in the fiscal policy in the two year interval and Ω𝑡𝑡 that the forecast is conditional to the information set available at the beginning of time period t. Since the change in the fiscal policy is estimated by using government fiscal balance, the positive values for Δ𝐹𝐹𝑖𝑖,𝑡𝑡:𝑡𝑡+1 indicate fiscal consolidation and negative values expansionary fiscal policy.
The data used by Blanchard and Leigh is from the IMF World Economic Outlook 2010 and 2012. They used real GDP growth and general government structural balance projections from the first issue of the IMF World Economic Outlook 2010 and compared the results with the realized real GDP growth available in the second issue of the World Economic Outlook 2012. Blanchard and Leigh used the most recent data available at the time they were writing their paper and they noted that some, especially their estimations for the most recent periods, were preliminary. They focused on the European Union member states but included also data from three non-European Union members identified as advance economies: Norway, Iceland and Switzerland. Due to unavailable data for the fiscal balance they dropped out Lithuania, Estonia, Luxembourg and Latvia. Thus, their data set consists from 26 European Economies.
Blanchard and Leigh found a statistically significant coefficient (-1.095) for their baseline regression which in the context of their setup suggests that the forecasters underestimated the effect of fiscal policy change on real economy. They tested their baseline results by using other vintages and found negative and significant relationship for projections made in 2009 (-0.667) and 2012 (-0.358). Interestingly they could not confirm
23
statistically significant results for forecasts made in 2011 even though the sign of the coefficient (-0.467) was as expected.
Blanchard and Leigh also tested their baseline regression for various variables and by using projections made by other institutions. They noted that when testing for omitted variables in the context of forecasts, using ex-post variables would not serve the purpose and would result in coefficients without clear economic interpretation. In other words, since the main interest is to investigate whether forecasters misestimated the impact of fiscal policy changes on growth, control variables are used to test if the forecast error could actually have been caused by some omitted variable. After running the regression with multiple control variables they concluded that their regression was not biased by any omitted variables. Furthermore, by running the regression with data from a non-crisis period, they estimated the coefficient (-0.077) to be close to zero and statistically insignificant. The result supports their hypothesis according to which the real GDP forecast errors are due to unpredictably large impacts of fiscal policy changes on real growth during recession which surprised the forecasters. Based on their analysis, Blanchard and Leigh suggest that the larger than expected fiscal multipliers tend to surprise forecasters especially at the beginning of a crisis. Furthermore, based on their estimates for different vintages, they suggest that either the size of the fiscal multipliers decrease along the crisis which leads to a decline in the regression coefficient or the decrease in the coefficient is due to a learning process. The latter would mean that the forecasters actually modify their models to fit the real data better and take higher fiscal multipliers better into account.
24
4.2 DATA AND METHODOLOGY
This thesis applies the methodology used by Blanchard and Leigh and extends their model to the Nordic banking crisis in the early 1990s. The underlying motivation is to investigate whether a systemic relationship between forecast errors and fiscal balances can be identified by using data from other crises and whether it is possible to find further support for the suggested larger fiscal multipliers. If the results are in line with Blanchard and Leigh’s, it can be suggested that the higher fiscal multipliers are not just a one-off event related to the recent crisis but that there are certain economic circumstances which make the economy more sensitive to fiscal policy changes. To make the results as comparable as possible, Blanchard and Leigh’s methodology is followed as closely as possible. The data used in the empirical section of this thesis is mainly from the OECD World Economic Outlooks which are published twice a year. The data is collected from the first issues of each year which provided estimations for the year the issue is published and for the year ahead. In other words, the projections for the two year intervals are from the beginning of the time period t as explained above. The data is collected for 22 years time period, 1990-2012, to ensure that non-crises periods are included as well in order to test the hypothesis of zero coefficients for non-crisis periods. To follow Blanchard and Leigh, the available data is collected for the members of European Union (Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Luxembourg, Netherlands, Poland, Portugal, Spain, Sweden and United Kingdom) and the European non-EU members which are characterized as advanced economies (Norway, Switzerland and Iceland). Overall, the data is collected for 21 economies.
Fiscal balance is usually measured as a percentage of nominal or real GDP which would make the variable sensitive for cyclical variations, hence an unsuitable and an inaccurate measure for fiscal policy analysis. This would be inconvenient because it is important to isolate discretionary fiscal policy changes as accurately as possible. If for example a government’s tax revenue decreases during a recession while an incline in unemployment increases transfer payments, the value of the government’s fiscal balance decreases even though it has not been effected by an exogenous policy change. To minimize the effect of cyclical variation of fiscal balance on regression output, cyclically adjusted fiscal balances are used. The cyclically adjusted fiscal balance is measured as a percentage of the trend output and it controls for the cyclical movements and separates the cyclical influences from non-cyclical variations. Hence, it can be seen as “a
cause rather than an effect of output fluctuations and may be interpreted as indicative of discretionary policy adjustments”, which makes it suitable for the purpose of this paper. (OECD, 2011) Blanchard and Leigh
controlled their baseline regression with data from other forecasting institutions, including OECD, to investigate whether the systemic relationship was caused by a misspecification of forecasting models in IMF or if other forecasters systematically misestimated the real GDP growth as well. They found support for the latter. When running the regression with the OECD data, Blanchard and Leigh used the underlying
government net lending as a measure of fiscal policy. The estimates are strongly based on cyclically adjusted
