Oil Prices, Inflation and the Monetary Transition Mechanism in Iceland
A vector autoregression analysisÞorsteinn Andri Haraldsson
in Partial Fulfilment of the Requirements for the Degree of Master of Science in Economics
Thesis Supervisor: Dr. Gerard Kuper
Faculty of Economics and Business
Oil Prices, Inflation and the Monetary Transition Mechanism in Iceland.
This dissertation equals 20 ECTS credits towards a partial fulfilment for the M.Sc. degree in Economics at the Faculty of Economics and Business.
© 2015, Þorsteinn Andri Haraldsson
No part of this dissertation can be reproduced without the full permission of the author.
Abstract
1
Introduction
Inflation creates uncertainty. This uncertainty can lead to various costs, especially in times of high inflation. A high rate of inflation means that the owners of capital demand higher interest rates on their savings to compensate for the decreasing value of their wealth in real terms. Menu costs of firms become more evident, and consumers get more distorted from the real value of the currency. This can be exploited by firms as consumers start having a harder time to distinguish between relative price changes and changes in the general price level. An inflationary turbulent environment may also lead to negative real interest rates, making lenders and financial institutions more reluctant to provide long-term loans/mortgages in the process. Furthermore, high inflation rates tend to increase the volatility of inflation. Finally, a high rate of inflation is believed to have a negative impact on economic growth, and considerable long-run effects on the standard of living (Barro, 1995; Jónsson, 1999; Pétursson, 2000b).
Abrupt changes in oil prices are generally considered to have significant effects on the economy, particularly on inflation and output. However, these effects are expected to differ between oil importing and oil exporting countries. Monetary authorities in oil importing countries should thus pay special notice to oil prices and take them into account when formulating their monetary policy. For the small, open, oil-importing economy of Iceland, maintaining a low and stable rate of inflation is of paramount importance. However, that has proven to be a hard task, as is evident from figure 1.
Figure 1: 12-month change in the consumer price index from 1969 to 2015 (Statistics Iceland, 2015b).
In recent times, crude oil price volatility has increased significantly. Oil prices remained quite stable from 1991 to late 1999, or around $20 per barrel. However, after 1999 the volatility increased notably, with prices peaking at roughly $150 in mid-2008. Only a year later, oil priced plummeted down to $35. After shooting back up and ranging from $100-$120 for almost four years, oil prices have once again taken a nosedive (EIA, 2015). Therefore, the relationship between oil prices, inflation, and monetary policy in Iceland is interesting to explore further.
The causal link between crude oil price shocks and higher inflation rates is straightforward. An increase in crude oil prices is usually followed by an increase in oil products used by consumers and firms, such as fuel and heating. As fuel is derived directly from crude oil, prices tend to change almost instantaneously relative to crude oil price changes. This leads to an increase in the energy cost component of the consumer price index (CPI) and consequently, inflation rates rise. In the case of Iceland, oil has on average represented 4.87% of the CPI since 2001, while oil consumption in the year of 2013 amounted to 640 kilotons. Thereof, fishing vessels, automobiles and airplanes were responsible for more than 95% of that consumption.1 The direct effect of crude oil price increases on inflation in Iceland should thus be considerable (Statistics Iceland, 2014). Cologni and Manera (2008) point out there could also be indirect effects on inflation, caused by firms and consumers. Firstly, higher fuel prices cause households to increase their expenses. In light of a temporary increase in oil prices consumers wishing to maintain their real purchases of fuel need to either decrease their savings, borrow, or cut down on other goods. Alternatively, consumers might respond to the higher cost of living by demanding higher wages. The higher wage demand could then be channelled to the CPI through firms raising prices due to increased wage costs. Secondly, higher oil prices increase production and transportation costs of firms, which in turn will respond to those higher costs by raising prices. As a result, there is an increase in inflation. Finally, sharp increases in crude oil prices spread through currency flows to the exchange rate, thereby depreciating the currency and increasing inflation. As such, monetary authorities with goals of maintaining a certain rate of inflation have to concern themselves with oil prices fluctuations.
1 Oil consumption has been rapidly decreasing as a proportion of total energy use in the last century due
In 2001 the Central Bank of Iceland stopped using the exchange rate as its main monetary policy instrument and adopted inflation targeting in order to increase price stability. Restraining inflation to a desired target is of great importance to the economy, as high inflation has had severe impacts on the economy in the past. Inflation-indexation of financial liabilities, most notably on household loans and mortgages, further complicates matters. Although the inflation-indexation creates stability in an inflationary turbulent environment in the short-run, the long-run effects of inflation spikes have been seen to have serious repercussions on households’ mortgages and wealth. Moreover, the inflation-indexation proved to severely restrict the Central Bank’s ability to influence lending by local banks in the period leading up to the economic crisis in 2008. To this day, the indexation in its current form is highly debated in Iceland, as it is unclear whether it benefits society in the long-run.
With that in mind, this thesis aims at studying the effects of exogenous oil price shocks, or innovations, on the Icelandic economy. Specifically, this thesis inspects the response of inflation, the real exchange rate of the króna (the local currency, ISK), and the policy rates of the Central Bank when faced with oil price shocks. If oil prices increase inflation, how does the Central Bank react? Does it have more operational space to control inflation to a greater extent? To examine this relationship, two different vector autoregressive (VAR) models were applied. The first model adopts a linear specification of oil prices, while the second model replaces the linear specification with a non-linear specification. The reason for examining both models is that the empirical literature on oil price shocks shifted from a linear specification of oil prices over to a non-linear specifications around the 1980’s, after the linear relationship between oil prices and economic activity began to lose significance. To the author’s knowledge, this is the first paper to address this relationship between oil prices, inflation and the monetary transition mechanism in Iceland by means of a VAR model.
shocks between countries (Cuñado & Pérez de Gracia, 2003, 2005; Jiménez-Rodríguez & Sánchez, 2005; Cologni & Manera, 2008). This indicates that there isn’t a one-for-all solution regarding oil price shocks, and therefore, this paper adds to the current literature by focusing exclusively on one, small oil importing country that hasn’t been covered before. In light of the importance for the Icelandic monetary authorities to maintain inflation at a moderate rate, the relevance of the topic should be of interest to both monetary authorities and policy makers in Iceland.
2
The Icelandic economy and monetary policy
In Iceland, inflation plays a critical role, probably more so than elsewhere. Historically speaking, inflation in the country has been very high and volatile. To counter the inflation and its drawbacks, inflation-indexation of financial liabilities was introduced in the 1979 by tying the consumer price index to liabilities.2 With this indexation the borrower paid a fixed real interest rate, as the nominal value of the liabilities increased (and decreased) in line with the inflation rate, and the lender received his money back in real terms, freeing him from the negative effects of inflation (Jónsson, 1999). Quickly after 1979, inflation-indexation of loans became a standard practice in the country as it mitigated the effects of inflation; the lender now faced less risk and the borrower’s monthly payment burden remained more stable (Jónsson, 2009). By the end of 2014, roughly 79% of total lending to households in Iceland embodied inflation-indexed loans and 71% of household mortgages were indexed (Central Bank of Iceland, 2015).
Inflation-indexation in general is not uncommon and can be found in most developed countries, for example where governments issue inflation-indexed bonds. However, inflation-indexation on mortgages is not as common and is usually only apparent in countries where inflation is highly unstable. In this context, Israel has also issued mortgages with similar characteristics to the ones in Iceland.3 Both these countries have experienced prolonged periods of high and volatile inflation (World Bank, 2014).
Taking into account how widespread inflation-indexation is in Iceland, it is easy to imagine that inflation spikes can have considerable impact on households’ debt and real wages. That is exactly what happened after the recession of 2008, when in a matter of three years the ratio of households that had a very hard time to make ends meet rose from 5.6% up to 13.7%, arrears on mortgages increased by 84%, and households’ real wages shrunk by 12%. Although there were other factors at play as well, such as increased
2 Originally, the indexation was achieved by tying liabilities to the so-called credit-terms index, which
was a weighted average of the CPI and building cost index. The wage index was incorporated in 1989, but in 1995 the credit-terms index was changed to reflect only the CPI with a one month lag. Nowadays, the indexation refers to the headline CPI (Jónsson, 1999).
unemployment, the rapidly rising inflation certainly played its part (Statistics Iceland, 2015c).
2.1 Empirical literature on oil price shocks and monetary decision
making
The empirical literature on the effects of oil price volatility on economic activity and monetary policy is quite extensive. Hamilton (1983) performs one of the first studies, where he investigates the linear relationship between oil price shocks and post-World War II economic recessions in the U.S. by means of a Granger-causality test. Hamilton concludes that the relationship appears to be too strong to be a simple coincidence. Numerous authors have built on the work of Hamilton. Burbidge and Harrison (1984) use a seven variable VAR analysis, and a study by Gisser and Goodwin (1986) inspects the relationship between nominal crude oil price and four indicators of macroeconomic performance in the United States. By means of a Granger-causality test the authors discover that all four indicators: real GDP, inflation, unemployment, and real investment were significantly affected by oil price shocks. Furthermore, Dotsey and Reid (1992) find out that the oil price shocks, along with monetary policy innovations (interest rate changes) are associated with movements in real economic activity. Rotemberg and Woodford (1996) show that it is easier to show the magnitude of output and real wage decline following an oil price innovation when the standard neoclassical growth model is assumed to have imperfect competition. Carruth, Hooker, and Oswald (1998) offer evidence that oil price innovations in 1990 were a determining factor in the recession in the U.S. following Iraq’s invastion of Kuwait. Finally, Hamilton (2003) extends his earlier work and provides evidence that the relationship between oil prices and output is not a statistical coincidence. However, Hamilton (2011) stresses that his earlier conclusion did not state that oil price shocks were the sole cause of post-war recessions, but points out that the oil shocks of the past decades certainly were a substantial contributing factor of at least some of the post-war recessions.
Hooker’s results and states that they are correct and undeniable. However, Hamilton proposes a new, non-linear, way to measure oil price changes where he defines the oil price explanatory variable as net oil price increase, or NOPI. The NOPI variable’s value in each period is defined as the amount that the (log of) oil price in the current period exceeds the maximum (log) price in the preceding four periods; and 0 otherwise. This new specification, the so called net specification, re-establishes a significant relationship between oil price increases and economic recessions in the U.S. Previously, two additional non-linear specifications for oil prices had been proposed: the asymmetric specification by Mork (1989), where increases and decreases in oil prices are considered as separate variables; and the scaled specification proposed by Lee et al. (1995), which takes into account the volatility of oil prices. Hamilton’s net specification is also asymmetric in that sense that it only captures shocks that increase oil prices, not oil price decreases.
The relationship between oil prices and monetary policy has been researched extensively. Tatom (1988) addresses the issue of whether oil price shocks have symmetric effects on macroeconomic variables. He concludes that prices, output, and production do indeed react symmetrically to oil price shocks when the stance of the monetary policy is taken into account, which suggests that the monetary authorities themselves behave asymmetrically to oil price shocks. Bernanke, Gertler and Watson (1997) develop a VAR-based method that decomposes the impact of a given exogenous shocks down to the direct effects of the shock itself, and the part that arises due to a shift in the monetary policy. By treating oil price shocks as exogenous, they find that the recessionary impact of oil price shocks are substantially enhanced when interest rates are raised to control inflation. However, when the monetary policy is left unchanged there is a positive effect on both real GDP and inflation. Their results therefore indicate that the monetary policy itself is the primary source of the economic downturn following an oil price shock. Barsky and Kilian (2001, 2004) also examine the role that monetary policy plays in the relationship between oil price increases and output. Their main findings complement those by Bernanke et al. (1997) and state that much of the oil price increases of 1973-1974 were caused by monetary expansion. This monetary expansion was then the driving factor behind the subsequent fall in output, not the oil price shock.
Hamilton and Herrera (2004) however challenge the results by Bernanke et al. (1997) on two grounds. Firstly, they questions whether the monetary authorities actually have the power to implement such a policy suggested by Bernanke et al. (1997), and particularly mention that the policy suggestions by Bernanke et al. (1997) are highly unrealistic. Secondly, they criticise the lag length used by Bernanke et al. (1997), which is considerably smaller than previous researchers have used. By using a more reasonable lag length of four, Hamilton and Herrera show that the monetary authorities efforts in adverting a recession would have been futile.
2.2 The Central Bank of Iceland in historical context
The Central Bank of Iceland was founded it 1961. With its foundation it was given all the most traditional central banking rights, such as the right to issue notes and coins, manage the foreign exchange rate reserves, the power to regulate interest rates in the economy, and to intervene to influence liquidity of the banking system. However, the formally independent central bank did not have full independency in its decision making as it was required by law to support the economic policy of the government. This restrained the Bank in making any major policy alterations, for example changing its policy rates, without the consent of the government. This arrangement did not produce any conflicts during the sixties, but started creating serious problems following the oil crisis in the seventies. With inflation rapidly accelerating, mainly due to increased oil prices between 1973 and 1974, the government nonetheless remained firm on maintaining the interest rates low, despite protests from the Central Bank. This led to a period of negative real interest rates and significant fall in financial savings. After the first oil price shock, negative external shocks started to appear at increased rate and a second oil price shock hit between 1979 and 1980. This further fed the already high inflation, which would become very persistent and would stay above 40% until 1984 (Andersen & Guðmundsson, 1998, Central Bank of Iceland, 2008).
During the seventies the exchange rate was used as the main policy instrument. The króna was initially pegged against the U.S. dollar, but later against various baskets of trading partner countries’ currencies. The policy rule of the Bank was to keep the nominal exchange rate fixed in upturns and devalue it in recessions. The devaluations stimulated the economy during recessions, but unfortunately increased inflation due to a depreciation of the currency. 4 At the same time, the depreciations cut down real wages, to little pleasure of the public. However, the real wage cuts offset the inflationary impact to some extent and thus prevented inflation to spiral out of control. Up till the late eighties, devaluation of the exchange rate would be frequently used as an instrument to boost economic activity. In the mid-seventies, Iceland also proved victorious over the U.K. in the Cod Wars, which
4 As most of the goods in the marine industry are exported, depreciating the króna increases the net trade
was ensued by growth in revenues from fisheries and a vast increase in investment in new vessels and facilities. In combination with the frequent devaluations of the króna, this resulted in overheating of the economy and inflation rates ranging around 30% to 50% during the seventies (Gudmundsson & Kristinsson, 1997; Jónsson, 2009).
The environment of a high, volatile, and persistent inflation with negative real interest rates led to the introduction of inflation-indexation of financial liabilities in 1979.5 The indexation was implemented by tying liabilities to the so-called credit-terms index (explained in a previous segment), and would become the first step out of three towards liberalisation of domestic interest rates. The indexation became an instant hit; it eradicated the problem of negative real interest rates and reversed the trend of deteriorating financial savings. In addition, the indexation opened up the path for long term lending, as lenders no longer had to worry about facing negative real interest rates. Banks were immediately authorised to index lending, while indexation of deposit accounts was carried out in 1980. Originally, the deposits were bound for minimum of two year, but that was reduced down to three months in 1982. Although real interest rates had finally become positive, inflation however kept on growing as the government still remained firm on maintaining low nominal interest rates. A fall in fish catches in 1982, along with the combination of indexation of wages and indexation of liabilities, ultimately created a spiral that shot inflation up to 103% in the early months of 1983. As the economy direly needed stabilisation measurements, wage indexation was suspended in May 1983.6 This action, along with a few others, proved to calm down inflation and eventually brought it down below 30% a year later.7 The inflation rate would remain at a similar rate for the next three years (Andersen & Guðmundsson, 1998; Central Bank of Iceland, 2008).
Between 1984 and 1986 the second step towards domestic liberalisation of interest rates was taken when commercial banks were given full liberty to specify their own deposit and lending rates, but prior to that the Central Bank had regulated all interbank interest rates.
5 Real interest rates had been around negative 20% in 1978 (Andersen & Guðmundsson, 1998).
6 This action was initially meant to be temporary, but was subsequently made permanent (Andersen &
Guðmundsson, 1998)
7 The other stabilisation measurements included placing ceilings on wage raises for the rest of 1983, using
During these years the exchange rate policy also became more restrictive, with exchange rate stability become a priority. The third and final step was the Interest Rate Act of 1987, which entitled full negotiating freedom for interest rates. This amendment abolished the rights of the Central Bank to control the interest rates set by domestic banks, and led to increased competition amongst the banks. The Central Bank still kept a close eye on money market interest rate and began using the rates on treasury bonds in the secondary market as its policy instrument to try to affect the market interest rate. However, it proved to be hard for the Bank to influence those rates due to the freedom given earlier to banks to determine their own rates. Furthermore, the money market itself had only just began to form and was at that time practically non-existent, which further limited the Bank’s ability to influence the market interest rates. The newfound competition amongst the banks eventually led interest rates to shoot up to 25% in 1988, with inflation following and reaching nearly 30% that same year (Gudmundsson & Kristinsson, 1997; Pétursson, 2001; Central Bank of Iceland, 2008).
Figure 2: Money market rates, policy rates (monthly) and inflation (12-month change) from 1987 to 2015 (Central Bank of Iceland, 2015; Statistics Iceland, 2015).
The evolution of money market interest rates, policy rates, and the inflation rate from 1987 to 2015 are plotted in figure 2.8 In 1989, the Central Bank once again decided to place more emphasis on using the exchange rate as its main policy instrument, rather than the interest rates, and now devoted itself to keep the exchange rate stable. That same year the
8 Both the repo rates and the collateral rates are depicted, as the collateral rates took over as the main
policy instrument in 2007. Section 4 will explain the difference between the two rates in greater detail.
0% 5% 10% 15% 20% 25% 30% 35% 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 Inflation Money Market rates Inflation target Collateral rates Repo rates
economy entered a deep recession after the cod stock collapsed due to severe overfishing. With one of the country’s main industries paralysed, it was decided to devalue the króna and freeze nominal wages in the country in 1990. This act managed to bring down inflation below double digits in late 1990, but came at the cost of nearly 20% real wages cut for the public. Devaluation of the króna was resorted to again, both in 1992 and 1993, in order to boost economic activity.9 The public grew unruly with the constant cuts in purchasing power and the demand to abandon the idea of becoming a Scandinavian-like welfare state became greater than ever. The focus was now turning towards a more free-market oriented economy which subsequently paved the way for the banking system to break free from its chains (Gudmundsson & Kristinsson, 1997; Pétursson, 2001; Jónsson, 2009).
Through the nineties, the financial liberalisation of the country kept on going and the money market started to take proper form. Accordingly, the Central Bank started to place more emphasis on regulating the market interest rates and changed its policy rates from being rates on treasury bonds over to rates on repurchase agreements (repos) between the Bank and financial institutions. The repo rates would remain as the key policy rates for the Central Bank until 2007, when they were correctly replaced by the rates on collateral loans from the Central Bank (Gudmundsson & Kristinsson, 1997; Central Bank of Iceland, 2007b).10
With enhanced focus on using the policy rate as a tool to stabilise the economy, inflation was finally managed to be brought down below 5% in 1992. This was a start of a period where inflation would finally remain stable, staying below 5% for almost a decade. The next act towards financial liberalisation was the liberalisation of long-term and short-term capital movements in 1994 and 1995, respectively. The final piece of the puzzle was then the privatisation of the main companies in the financial industry. During these times Iceland would become one of the leading free-market oriented economies in Scandinavia (Gudmundsson & Kristinsson, 1997; Jónsson, 2009).
9 These decisions to depreciate the króna did not come from the Central Bank, but came as a request from
the local government (Gudmundsson & Kristinsson, 1997).
10 The repos were denoted in annual rate of return instead of nominal interest rates. The rates on collateral
With the events surrounding the dot-com collapse and 9/11, inflation slowly started to take flight again along with the interest rates. Following growing pressure on the Central Bank to abandon the exchange rate policy, adjustments were made in 2001 and an inflation target was adopted as the new monetary policy. Along with this major policy reform, the króna, which had previously been fixed to a basket of currencies of the main trade partners of Iceland, was floated in 2001. This action led to inflation firing up to roughly 10%. The Central Bank nevertheless restrained itself and kept its policy rates stable. Eventually, the inflation rate started to subdue in mid-2002 and gradually fell down to the desired target, along with a steady decrease in the policy rates themselves (Jónsson, 2009, Central Bank of Iceland, 2015).
The privatisation of the formerly state-owned banks started with a partial privatisation in 1998, and full privatisation was finally reached in 2003. Before the privatisation of the state-owned banks, the assets of the banking system as a whole had been around 97% of the financial system and the equity ratio around 7.3% of the country’s GDP. After the privatisation, the wheels started to spin in the economy at a whole new level and the three big banks, namely Íslandsbanki (later Glitnir), Landsbanki and Kaupthing, began to expand internationally.11 However, in order for the international expansion to continue, the banks had to find new ways to access liquidity as the domestic deposit base was too small to finance the expansion. The banks did so by tapping into the apparently endless liquidity pool of foreign wholesale markets. This allowed them to enlarge their balance sheets. Another international ventures that helped the banks access more credit was the foundation of foreign online-only deposits accounts, namely ICESAVE in the U.K. and the Netherlands, and Kaupthing Edge in Sweden, Finland, U.K., and Germany. By having no retail branches, the banks argued that they could provide higher interest rates on deposits than competitors. This operation proved to be a success and deposits streamed into the ICESAVE accounts when they were opened for services in the U.K. in 2006. However, there were two vital differences between the accounts from Landsbankinn and Kaupthing; the ICESAVE accounts were operated under a guarantee from the Icelandic government, while the Edge accounts were under protection of the respective host country. This difference would prove to have grave consequences. By the September 2008 the banking
11 For a detailed and comprehensive description of the chronicles of the Icelandic banking sector and the
system´s assets would total more than elven times the Icelandic GDP. Along with the rapid growth of the banks, GDP per capita in Iceland raced up and by the end of 2007 Icelanders ranked number eight amongst OECD countries in terms of wealth (Jónsson, 2009; Statistics Iceland, 2015a).
With the rise of the banking sector and the prosperous economic period that followed, a housing bubble started to form. In a matter of five years, total lending to household increased by 159%. In order to try to influence lending by the banks, the Central Bank started to gradually raise its policy rate. However, the inflation-indexation proved to severely restrict the Bank’s ability to influence the lending. As rising housing prices fed inflation (housing costs had about 20% weight in the CPI), housing prices themselves further rose due to the inflation-indexation, and the Central Bank had to keep raising the interest rates. This spiral would send the policy rate of the Central Bank all the way up to 15% during the peak of the boom period (Jónsson, 2009).12
The high interest rates set by the Central Bank eventually started to attract foreign investors, and the króna appreciated to levels it had not seen for many years. This high exchange rate however also lured interest from foreign hedge funds, who thought the króna was overvalued and aggressively started to short it in 2006. The króna took a sharp dive in the process and inflation rose. After fighting back the hedge funds in numerous ways, Iceland soon recovered and the króna was brought to a stable level along with inflation. However, the banks, who had kept a relatively low profile in the international media, were now on everybody’s radar. More importantly they were in a fragile state (Jónsson, 2009).13
Following the impending collapse of Bear Stearns (a U.S. based global investment bank) in the middle of 2007, more woes were brought to the Icelandic economy. The amounting financial distress in the world did not only lead hedge funds to start shorting
12 In 2008 the banks were also offering up to 15% interests on deposits in savings accounts (Jónsson,
2009).
13 This was most apparently seen by looking at the credit default swaps (CDS) spreads of the banks. A
the króna again, but now also stocks in the three major banks. Despite that, the banks managed to operate surprisingly well for the next year, right until the international liquidity crisis hit in August 2008. With the bankruptcy of Lehman Brothers (a U.S. based international financial services firm) a month later, things finally started to crumble. News started to spread in September that the Icelandic banks were in serious trouble and a bank run began, both on the ICESAVE accounts in the U.K. and the Netherlands, and in Iceland.14 The run on the government guaranteed ICESAVE accounts was particularly serious, since deposits on them amounted to nearly 70% of the Icelandic GDP at the time. With inflation on the rise and the Central Bank eager to get it down to the desired target, the policy rates climbed up to 15% during final days before the meltdown and nominal wages were frozen as a stabilisation measurement, just like in 1989. The collapse of the banking system finally came about in October 2008 (Jónsson, 2009).
It was announced on 29 September 2008 that Glitnir would be nationalised in an effort to restore faith back into the financial system. The Central Bank decided to lower its policy rate down to 12% in an attempt to get some life back into the frozen economy, but after that effort proved fruitless it quickly raised the interest rates up to 18% as capital was flying out of the country. Ultimately, all official currency exchange was suspended on 6 October and both Glitnir and Landsbankinn were taken over by the Icelandic FSA (Financial Supervisory Authority) on 7 October. For a while it looked like Kaupthing was going to manage and ride out the storm, but after a critical and unexpected intervention by the British government on 8 October the fate of Kaupthing was also sealed.15 At midnight that same day, executives at Kaupthing finally gave up and the bank was nationalised on 9
14 I myself clearly remember this event. I was sitting in the classroom when a fellow student came running
into the room, shouting that everybody should run to the nearest bank to take out his saving; the banks were defaulting! Few students followed the advice and ran to the nearest bank, where retirees could be seen walking out with a bag full of savings.
15 The British government imposed an antiterrorism law that authorised them to seize all assets of
October (Jónsson, 2009).16 On top of everything, a roughly 80% depreciation of the króna meant that most companies in the country were technically bankrupt, owing to the fact that about 70% of corporate debt in Iceland was in foreign currency. Eventually, strict capital controls were introduced in November 2008 in order to further prevent capital outflow and help stabilise the króna (Jónsson, 2009; Saborowski et al., 2014).17
One month after the events in October 2008, Iceland received a sovereign debt package from the IMF (International Monetary Fund) and the Nordic countries. However, the next two years would become very hard on the Icelandic economy, with unemployment reaching unprecedented heights and household debt ballooning due to inflation-indexation (Statistics Iceland, 2015a).
The economy finally started to show signs of recovery in 2011. Today, the economy has recovered marvellously; there has been GDP growth for the past four years, unemployment equals 4%, the loan from the IMF is projected to be paid up in 2015-2016, and the Nordic loan is already retired. Most importantly, the Central bank has managed to control inflation. The inflation rate was first brought down below the inflation target in 2009, and then again in 2014, where it has remained for twelve consecutive months. (Central Bank of Iceland, 2014c, 2014d; Statistics Iceland, 2015a).
2.3 The current role of the Central Bank of Iceland
Today, the chief objective of the Central Bank of Iceland is to maintain price stability, i.e. a low and stable rate of inflation. Price stability according to the Central Bank is defined as no more than a 2.5% (±1.5%) annual rate of inflation, which is based on a 12-month changes in the headline Consumer Price Index (CPI).18 Using the headline CPI is appropriate for numerous reason, but mostly due to the fact that the indexation of financial liabilities in the country is based on that measurement, as well as being the measurement that the Icelandic public is most familiar with (Central Bank of Iceland, 2001). The main
16 For detailed explanation of what happened to the deposits accounts in the U.K. and the Netherlands,
see Jónsson (2009), pp: 178-188.
17 The capital controls are still in place, but actions in order for their removal are already under way. On
April 10 the Icelandic Prime Minister, Sigmundur Davíð Gunnlaugsson, announced that a bill regarding the removal of the capital controls would be represented to parliament before its summer closing.
18 The headline CPI is measured monthly and consists of over 20,000 prices of more than 4,000 different
reason for setting a relatively high deviation rate from the desired target relies on the fact that the Icelandic economy is a small open economy, vulnerable to frequent inflationary trade shocks. Imposing and maintaining a lower target might prove to be unfeasible. If inflation deviates by more than 1.5% in either direction from the desired target, the CBI is required to hand in an explicit report to the Icelandic Government, explaining possible reasons for the deviation. This report should also indicate how the Central Bank will respond to the deviation and when it expects the inflation target to be reached again (Central Bank of Iceland, 2014a).
The inflation target policy was introduced in March 2001, but New Zealand had been the first country to implement such a policy in early 1990. By adopting the inflation target as a policy rule the Central Bank argued that it would achieve greater division from local policy makers, avoiding government intervention and become a more independent unit, therefore strengthening its credibility and bringing its operations more in line with international best practice. In return for this increased independency, the Bank would provide more transparency in its operations. As covered in the previous section, various types of fixed exchange rate policies had been in effect prior to the inflation target. Those policies are considered “hard” as the Central Bank was obliged to maintain the exchange rate of the króna within a certain boundary at all times. The inflation target is on the contrary considered “soft” as the Bank doesn’t strictly need to keep inflation within the tolerance level at all times. In addition to this fundamental difference between the two approaches, the inflation target is also believed to combine two aspects considered important for monetary policy. Firstly, it provides a reliable medium-term anchor for inflation expectations, and secondly, it allows for enough policy flexibility to respond to short-run shocks without endangering the credibility of the policy’s framework (Central Bank of Iceland, 2001; Pétursson, 2004).
The main instrument that the Central Bank uses to carry out its monetary policy is the interest rates on deposits at the bank or loans from the bank.19 By altering these so called
19 The Central Bank offers two types of facilities to financial institutions: standing facilities, which can
policy rates, the Central Bank in turn affects the money market interest rates and those set by financial institutions. Furthermore, the Central Bank can influence domestic inflation through the exchange rate. As money market interest rates have a strong impact on currency flows and thus on the exchange rate and inflation, the Bank can buy or sell currency in the interbank market and subsequently influence the exchange rate. However, foreign exchange market intervention is only executed if the Central Bank sees it necessary in order to maintain its inflation target or if it estimates that future exchange rate fluctuations could be a potential treat to financial stability (Central Bank of Iceland, 2014a).
As the fundamental objective of the Central Bank is to maintain price stability, the Bank does not concern itself with other economic targets, such as unemployment or balance of current accounts, unless those concerns are in line with the Bank’s inflation target. This argument is based on the idea that monetary policy can only influence nominal values of variables in the long-run, not real values. Nevertheless, the Central Bank is obliged to promote the government’s policy, but only to the extent that the policy doesn’t conflict with the inflation target (Central Bank of Iceland, 2001).
2.4 Inflation-indexation of households´ loans
Up until mid-2004, the state-owned Housing Financing Fund (HF Fund) was almost the sole provider of mortgages in the country. The HF Fund operated under government guarantee and exclusively issued inflation-index mortgages.20 In the autumn months later that year the banking sector, namely Kaupthing, started to provide mortgages at a more competitive rate. In line with the increased competition, and after a fundamental change in the issuing of mortgages by the HF Fund, the HF Fund lowered its interest rates from 5.1% down to 4.15%. The banks soon followed and lowered their interest rates to the same level, but also offered the consumers the appealing option of being able to get up to 80% financing. That vastly increased their market share of mortgages (Central Bank of Iceland, 2007a; Jónsson, 2009).
Data on lending to Icelandic households has been recorded since 1991. Total lending to households can be seen in figure 3, while figure 4 demonstrates what type of mortgages
have taken.21 Since 1997, inflation-indexed loans have represented 71%-89% of total lending and in 2007 they represented 93% of mortgages. Unindexed mortgages did not become available at all until late 2008 when the banking sector started to provide them. By offering unindexed mortgages, the banks were reacting to increased demand for alternatives to the inflation-indexed loans and the foreign currency-linked loans that had become popular around 2005. In the end of 2014, household debt amounted for 1,765 billion króna, or 86% of the national GDP. Thereof, the banking sector and the HF Fund act as the main providers, representing 43% and 47%, respectively. The remaining 10% are provided by pension funds and insurance companies. Bearing in mind that 79% of the current outstanding liabilities is inflation-indexed, it is easy to imagine that inflation shocks can affect the household debt burden of the economy (Central Bank of Iceland, 2015).
Figure 3: Lending to households by type of loan, in million króna (Central Bank of Iceland, 2015).
21 This data constitutes all loans to households, not only mortgages although they do represent the
substantial part of the data. Data solely on mortgages didn´t start to get recorded until early 2007.
0 200.000 400.000 600.000 800.000 1.000.000 1.200.000 1.400.000 1.600.000 1.800.000 2.000.000 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Foreign currency-linked Indexed Unindexed Leasing contracts
Figure 4: Type of mortgage, as a proportion of the total (Central Bank of Iceland, 2015).
Before the recession of 2008, inflation-indexation had generally been regarded as a success story in Iceland and represented 93% of household mortgages in mid-2007 (foreign-currency linked mortgages represented the remaining 7%). The fundamental difference between indexed and unindexed mortgages is that the indexed borrower does not bear the whole cost of inflation immediately. In times of high inflation, the repayments on an indexed mortgage stay relatively stable as only pays a part of the cost of indexation and the rest is transferred back to the principal. On the contrary, repayments of an unindexed mortgage with flexible interest rates might become insurmountable as the flexible interest rates are based on, amongst other things, the policy rate of the Central Bank and interest rates on deposits. When inflation shoots up, these rates usually follow, along with a substantial short-term increase in the interest payments on the mortgage. As inflation has historically been very volatile in Iceland, the bumper provided by the indexation can therefore become very valuable. Nevertheless, the bumper has to be viewed as a temporary solution since the indexation increases the principal of the loan in the long-run. Thus, it is up for debate whether the value of having the bumper outweighs the increased cost in the long-run.
The inflation spike during the recession of 2008 sparked new debates regarding the negative effects of inflation-indexation. In light of the ballooned household debt following the crisis, the government executed a major economic intervention in 2014, titled
Leiðréttingin (translated as The Adjustment), in order to reduce this debt. This government
intervention constituted a transfer of 150 billion króna from the government coffers to the 56,000 households’ who had undertaken an inflation-indexed mortgage between 2004 and
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 7.2007 5.2008 3.2009 1.2010 11.2010 9.2011 7.2012 5.2013 3.2014 1.2015
2007 (Prime Minister’s Office, 2013).22 In a working paper released by the Central Bank in 2012, this action was highly criticised.23 Furthermore, the negative debates surrounding inflation-indexed mortgages can clearly be seen when inspecting statistics on mortgages to households. Once dominant on the market, inflation-indexed mortgages have now shrunk to 71% of the total, while unindexed mortgages have grown to a market share of 29% (Central Bank of Iceland, 2015). Considering that unindexed mortgages have only been available for seven years, this has to be regarded as a dramatic reconstruction of mortgage lending in the country.
To show the impact of an inflation shock to an inflation-indexed loan, table A1 in appendix A shows the evolution of two inflation-indexed mortgages taken in 2007; one with an estimated inflation of 2.5% and the other with available observed inflation and 2.5% estimated inflation after that until maturity.24 This particular period is chosen to showcase the effects that the inflation spike in 2008 had on mortgages taken during the housing bubble. The difference between repayments on the mortgages is striking; the inflation spike results in a 25% increase in total repayments over the course of the mortgage! In light of the drastic measurements that have been carried out after severe inflation spikes, such as Leiðréttiningin, these results support the view that inflationary oil price increases should be of concern to the Icelandic monetary authorities who are devoted to price stability.
22 This reform has been highly criticised, especially from the younger generation who feels like they were
done a great injustice. Their argument is based on the fact that a transfer of a huge amount of government money to one social group exclusively results in worse state for the younger generations in the future, since the government could have used that money to either lower taxes, investment in health care, or lower the government’s monthly interest burden.
23 See Ólafsson and Vignisdóttir (2012).
24 Both loans have a principal of 29.8 million ISK, monthly repayments, mature in 40 years, and have
3
The VAR framework
To analyse the role of shocks in the economy, vector autoregression (VAR) methods were applied in this study. The VAR approach was introduced by Sims (1980) after his forceful critique on simultaneous equation models. Sims main criticism on simultaneous equations models revolved around the identification process and the often highly unrealistic exogeneity assumptions of variables. Both these steps involve many arbitrary decisions (Lütkepohl, 2005).
The VAR approach, which is a multivariate extension of a univariate AR model, consists of regressing each variable in the model on its own lagged values along with the lagged values of all other variables in the model. In other words, the VAR approach explains the endogenous variables of the model solely by their own history, apart from deterministic regressors (if necessary). This way each equation in the VAR model contains the same set of determining variables and can be measured consistently by ordinary least squares (OLS) methods (Gottschalk, 2010).
Impulse response functions and variance decomposition were then employed to analyse the relationship between variables in the model. The impulse response functions trace out the responses of current and future values of variables to a one standard deviation shock to another variable in the system. In order to get economically meaningful impulse responses, one needs to assume orthogonality among the shocks. When forecasting the endogenous variable over time the variance decomposition determines the proportion of the variability of the errors that is due to the variability in the structural innovations of all variables over time. The variance decomposition allows us to inspect the relative importance of one variable on the volatility of another (Stock & Watson, 2001; Gottschalk, 2010).
3.1 The VAR model
Following Lütkepohl and Krätzig (2004), the reduced form of a k-dimensional VAR(p) (i.e. with p lags) can be written as follows,
𝑦𝑡 = Φ𝐷𝑡+ 𝐴1𝑦𝑡−1+ ⋯ + 𝐴𝑝𝑦𝑡−𝑝 + 𝑒𝑡 (1)
where 𝑦𝑡 is a (k × 1) vector of time series variables, 𝐷𝑡 is the deterministic terms (intercept, trend), 𝐴𝑖’s (𝑖 = 1, … , 𝑝) are (k × k) coefficient matrices and 𝑒𝑡 is (k × 1) unobservable disturbance term. The error term is usually assumed to be zero mean white noise process with covariance matrix 𝐸(𝑒𝑡𝑒𝑡′) = ∑ .
However, without restrictions on the model the structural parameters in the reduced form of the k-dimensional VAR in equation (1) are considered unidentified. This identification problem can be demonstrated in a few simple steps. Consider the following simplified structural model,
𝛤𝑌𝑡 = 𝐵(𝐿)𝑌𝑡+ 𝑢𝑡 (2)
where Γ is a (k × k) matrix of structural parameters, 𝑌𝑡 is a (k × 1) vector of endogenous variables, 𝐵(𝐿) is the lag polynomial,25 and 𝑢
𝑡 is the vector of structural innovations. The reduced form of the model can be derived by premultiplying with Γ−1,
𝑌𝑡 = 𝐵∗(𝐿)𝑌
𝑡+ 𝑒𝑡 (3)
Here 𝐵∗ = 𝛤−1𝐵 and 𝑒
𝑡 = 𝛤−1𝑢𝑡. However, the model in equation (3) is not unique and therefore one cannot directly estimate it in hope to derive the “true” parameter values for Γ and B. Without imposing restrictions on the structural model it is considered unidentified (Gottschalk, 2010).26
3.2 Identification of the model
Following the model represented in equation (3), the intuition behind the identification of the model can be explained by starting to compute the Moving Average (MA) representation. By rearranging terms the MA form is acquired,
𝑌𝑡 = (𝐼 − 𝐵∗(𝐿))−1𝑒
𝑡 (4)
𝑌𝑡= 𝐶(𝐿)𝑒𝑡 (5)
where 𝐶(𝐿) = (𝐼 − 𝐵∗(𝐿))−1. This rearranging of terms is convenient, as the model in equation (5) represents the endogenous variables as a function as of the past and current reduced form innovations, while the model in equation (3) portrays the endogenous variables as a function of past values of themselves. Further rearranging of terms allows us to get economic meaning into the impulse responses generated from the model, as the model in equation (5) only depicts the impulse responses as a function of the reduced form innovations. Decomposing the model into
𝑌𝑡= 𝐶(𝐿)𝛤−1𝛤𝑒
𝑡, and (6)
25 𝐵(𝐿) = 𝐵
1𝐿 + 𝐵2𝐿2+ ⋯ + 𝐵𝑘𝐿𝑘.
𝑌𝑡 = 𝐶∗(𝐿)𝑢
𝑡, (7)
with 𝐶∗(𝐿) = 𝐶(𝐿)𝛤−1 and 𝛤𝑒
𝑡 = 𝑢𝑡, we now have the impulse responses in relation to structural innovations instead of reduced form innovations. To identify the model we need to know the matrix 𝛤 in order to compute 𝐶∗ and get meaningful impulse responses.
The first identifying restriction, which distinguishes VAR models from dynamic simultaneous equations, is the assumption of orthogonality amongst structural innovations 𝑢𝑡. By assuming that the structural innovations are uncorrelated (orthogonal) we get one non-linear restriction on 𝛤, providing us with the first identifying restriction. This assumption means that the structural variance-covariance matrix ∑𝑒 will take the following form: ( 𝜎𝑢1 0 ⋯ 0 0 𝜎𝑢2 ⋯ 0 ⋮ ⋮ ⋱ ⋮ 0 0 ⋯ 𝜎𝑢𝑘) (8)
The next step is normalisation; where the diagonal elements of ∑ 𝑒 , i.e. the variance of each variable, are set to one. This corresponds to a one standard deviation shock in the structural innovations, and indicates that ∑ = 𝐼𝑒 . However, this normalisation is only scaling and does not alter the model substantially. Now we can write the relation between the structural innovations and the reduced form innovations in the following way,
( 1 𝛾12 ⋯ 𝛾1𝑘 𝛾21 1 ⋯ 𝛾2𝑘 ⋮ ⋮ ⋱ ⋮ 𝛾𝑘1 𝛾𝑘2 ⋯ 1 ) ( 𝑒1 𝑒2 ⋮ 𝑒𝑘 ) = ( 1 0 ⋯ 0 0 1 ⋯ 0 ⋮ ⋮ ⋱ ⋮ 0 0 ⋯ 1 ) ( 𝑢1 𝑢2 ⋮ 𝑢𝑘 ) (9)
The final step in the identification process is to set restrictions on the parameters of 𝛤. This can be done either in a recursive order (Cholesky decomposition) or in a structural way, hence the term structural VAR (SVAR). These restrictions are derived from economic theory and to identify the model we need at least 𝑘(𝑘 − 1)/2 restrictions.
the cointegrating relationship into the model. By using this cointegrated VAR model, the non-stationary variables enter the model in levels. (Lütkepohl & Krätzig, 2004). However, Sims, Stock and Watson (1990) show that transforming models to a stationary form by first differencing or using cointegration when the variables are integrated is often unnecessary.
3.3 A VEC model
The ensuing section follows Lütkepohl and Krätzig (2004). The structural VEC model picks up from the reduced form k-dimensional VAR(p) model in equation (1), which is considered stable if
det (𝐼𝑡− 𝐴1𝑧 − ⋯ 𝐴𝑝𝑧𝑝) ≠ 0 for |𝑧| ≤ 1, (10) i.e. that the polynomial defined by the determinant of the autoregressive operator has no roots inside or on the complex unit circle, otherwise it is considered to have unit roots. If the polynomial in equation (12) has unit roots, then some or all of the variables are considered to be integrated. If the variables in the underlying model share a common stochastic trend they may be cointegrated, i.e. that there is a linear combination between the 𝐼(1) and 𝐼(0) variables.27 If the variables are cointegrated a VAR model is no longer the most appropriate model, as it does not account for cointegration. A vector error correction model, or VECM, builds the cointegrating relationship into the model. To build in the cointegrating relation, equation (1) can be written in VECM form written as
∆𝑦𝑡= Φ𝐷𝑡+ 𝛱𝑦𝑡−1+ Γ1∆𝑦𝑡−1… + Γ𝑝−1∆𝑦𝑡−𝑝+1+ 𝑢𝑡 (11) Here 𝑦𝑡−1 has been subtracted from both sides of the model in (1). The signs Π and Γ represent Π = 𝐴1+ ⋯ + 𝐴𝑝− 𝐼𝑘 and Γ =−(𝐴𝑖+1+ ⋯ + 𝐴𝑝) for 𝑖 = 1, … , 𝑝 − 1. Assuming that all variables are at most 𝐼(1), ∆𝑦𝑡 will not contain stochastic trends and the term 𝛱𝑦𝑡−1 is the only one that includes 𝐼(1) variables. Therefore, 𝛱𝑦𝑡−1 must also be 𝐼(0) and consequently contains the cointegrating relationship.
4
Description of data and tests
In the construction of a VAR model, four key decisions have to be made. The first decision is the choice of variables to include in the model. The second decision is related to the choice of lags to include in the model, as that can alter the impulse responses quite significantly. The third decision revolves around the treatment of non-stationarity in the data, like discussed in section 3.2, and the fourth and final decision relates to the identification of shocks.
The variables used in the model are the Consumer Price Index (𝑝𝑡), the Central Bank’s policy rates (𝑟𝑡), real exchange rate of the króna (𝑒𝑡),28 and the price of crude oil (Europe Brent price) in terms of U.S. dollar (𝑜𝑡).29 These four variables were carefully chosen to represent the monetary transition mechanism in Iceland.
It was decided to use monthly data for the period 2001M03 – 2015M02, which translates to 169 observations. The main reason to start with the March 2001, was the implementation of the inflation target. Prior to the inflation target the Central Bank mainly influenced the economy through exchange rate interventions, while using its policy rates to affect the short term market rates. After the implementation, the policy rates were instead strictly used to steer inflation. The policy rate therefore served a fundamentally different purpose before and after the inflation target was put into practice. As a result, the aforementioned time period was chosen to avoid possible distortions in the data due to structural changes.
The specification of most of the variables is quite self-explanatory. However, the specification of the Central Bank’s policy rates requires further elaboration. On 16 May 2007, the Central Bank redefined the rules and definition of its policy rates. Previously, the rates on repurchase agreements (repos) had been regarded as the Central Bank’s key policy rates. However, these repo rates did not actually involve any repurchase agreements, but rather worked as a loan against a collateral. As a result, these rates were redefined as rates on collateral loans. Another change that was implemented was the numerical presentation of the policy interest rates. Before, the interest rates were defined as the
annual rate of return and not the nominal interest rates. This was not in line with common
28 Real exchange rate of the króna is defined such that an increase indicates a real appreciation of the
króna.
practice, as it made the policy rates of the Central Bank of Iceland appear to be higher than they actually were. For example, the annual rate was 14.25% at the time of change, which translated to a nominal rate of 13.31%. Due to this discrepancy, the presentation of the rates was changed effective immediately (Central Bank of Iceland, 2007b). The policy instruments of the Central Bank were once again altered in May 2014, now by adding seven-day term deposits and discontinuing auctions of 28-day term deposits. This was done in order to improve the effectiveness of the Central Bank’s liquidity management, as well as prepare the Bank for the forthcoming liberalisation of capital controls. In 2015, the Bank started to refer to the seven-day term deposits as its key policy rate instrument, and those rates should therefore be the monetary policy rates used in this research. However, due to the limited timespan of the seven-day term deposits rate, the seven-day collateral rate will be used as the policy rates (Central Bank of Iceland, 2014b; 2015).
As discussed in section 2.1, the linear relationship between oil prices and economic activity began to lose significance in the 1980’s. However, when transformed to a non-linear specification, the significance of this relationship was re-established. Therefore, following Hamilton’s (1996) net specification of oil prices, a 𝑁𝑂𝑃𝐼𝑡 explanatory variable was constructed to estimate an alternative model. In this model the 𝑁𝑂𝑃𝐼𝑡 variable replaced the formerly linearly specified oil price variable. Although Hamilton’s specification was based on quarterly data, the same specification was be applied in this paper by including the previous four months. The 𝑁𝑂𝑃𝐼𝑡 variable was defined as:
𝑁𝑂𝑃𝐼𝑡= max{0, ln (𝑜)𝑡− ln (max{𝑜𝑡−1, 𝑜𝑡−2, 𝑜𝑡−3, 𝑜𝑡−4})}
All series were available in levels except for the policy rates, which are represented in percentages. The variables represented in levels were transformed to natural logarithms. Econometric analysis generally requires stationarity, and to check for stationarity of the model each variable was tested for unit roots. The Augmented Dickey-Fuller test (ADF) and the Phillips-Perron (PP) test were applied, and both indicated that all variables were I(1).30
The second decision, i.e. the selection of the appropriate lag length, is of critical importance, as pointed out by Hamilton and Herrera (2004) in their criticism on Bernanke
30 The outcome of both tests can be found in appendix C, along with other data description. The
et al. (1997). As such, four different criteria were considered to evaluate the appropriate
number lags to include, namely the Likelihood ratio test, the Akaike, Schwarz, and Hannan-Quinn information criterion (LR, AIC, SC and HQ, respectively). Moreover, each criterion was examined with several maximum lag lengths for robustness. When different optimal lag length were suggested, the Likelihood ratio test was used as the benchmark, with the other three criteria serving as a sensitivity analysis. The suggested lag lengths of the LR test, AIC, SC, and HQ are depicted in table 2. The LR test suggested a border case when 12 maximum lags were included, but otherwise it recommended 8 lags. The AIC criterion supported the LR test and also suggested 8 lags as the optimal choice. However, both SC and HQ criteria consistently suggested an optimal lag length of 2. For the empirical analysis it was decided to choose 8 lags, since that was the most frequently suggested lag length by the LR test and AIC criterion.31
Table 2: Optimal lag order selected by criterion, for an unrestricted VAR in levels.
Lags included LR AIC SC HQ
8 8 lags 8 lags 2 lags 2 lags
10 8 lags 8 lags 2 lags 2 lags
12 12 lags 8 lags 2 lags 2 lags
14 8 lags 8 lags 2 lags 2 lags
16 8 lags 8 lags 2 lags 2 lags
18 8 lags 8 lags 2 lags 2 lags
20 8 lags 8 lags 2 lags 2 lags
Before deciding on how treat the non-stationarity in the data, a Johansen test was ran to check for cointegration. The Johansen test indicated that there was one cointegrating relationship among the variables at a 5% significance level. Therefore, it was decided to apply a VECM to analyse the data. In the VECM, variables enter the model in levels instead of first differences to avoid loss of level information. The cointegrating equation from the VECM with 8 lags can be seen in equation (12).
𝑝𝑡− 1.210𝑜𝑡− 5.252𝑒𝑡+ 16.374𝑟𝑡= 0 (12)
31 For robustness, the empirical model was also estimated with both 2 and 12 lags. The IRF from those
However, this paper will not focus on the long-run relationship and only uses impulse response function and variance decomposition to interpret the results.32
4.1 The macroeconomic model and identification scheme
The fourth decision and last decisions in the construction of the VAR regards the identification scheme. The model contains four variables, and following section 3.2 it is clear that we need six identifying restrictions to fully identify the model. The identification scheme is acquired by utilising the institutional knowledge of the Central Bank of Iceland and economic theory.
Iceland’s consumption of crude oil products is solely provided by imported oil. Being a price taker on the market for oil, it is easy to accept that the country’s consumption has no effect on world price. This leaves us with the following equation:
𝛾11𝑒𝑜 = 𝑢𝑜 (13)
Here we have imposed three restrictions, 𝛾12 = 𝛾13 = 𝛾14 = 0.
The real exchange rate is assumed to be only determined by oil price shocks. As fuel consumption in the country is driven by imports, oil price fluctuations are assumed to have a contemporaneous effect on the exchange rate. Furthermore, exchange rate interventions by the Central Bank are executed when the Bank sees it necessary to maintain the inflation target or if future exchange rate fluctuation could endanger financial stability. As it concerns itself with oil prices when formulating its policy rate, and provided that oil price innovations induce inflation, the Bank might execute an exchange rate intervention
within-the-month in the light of oil price shocks. This leaves us with the following equation:
𝛾21𝑒𝑒+ 𝛾22𝑒𝑜 = 𝑢𝑒 (14)
Two restrictions have been imposed here, 𝛾23 = 𝛾24 =0. One might argue that the policy rate setting should affect the exchange rate, as increased interest rates should attract foreign investors. However, the assumption here is that interest rate settings only affect the nominal exchange rate, not the real exchange rate.
32 To inspect the point made by Sims et al (1990), i.e. that transforming models to a stationary form by
From the discussion in section 2, it is reasonable to assume that oil prices fluctuations as well as real exchange rate fluctuations have contemporaneous effects on the CPI.
𝛾31𝑒𝑜+ 𝛾32𝑒𝑒+ 𝛾33𝑒𝑝 = 𝑢𝑝 (15)
Here, one restriction has been imposed on the equation, 𝛾34 = 0. This restriction implies that inflation is not contemporaneously affected by the policy rates and is based on the assumption that inflation does not react within-the-month to policy rate changes. Bernanke
et al. (1997) adopt a similar restriction, based on the assumption that interest rates do not
affect the “slow-moving” variables such as output and prices within-the-month. This is a plausible restriction, due to the existence of planning and production lags.
Policy rate decisions by the Central Bank follow shifts in the inflation rate. Assuming that there are no information delays and that the Central Bank has instruments that allow him to predict the macroeconomic business cycle, the Bank is able to respond to inflation changes within-the-month. The Central Bank uses its own inflation forecast to set the current policy rates, so this seems reasonable. If the inflation rate is above the desired target, set by the Bank, the policy rates are expected to increase. Furthermore, policy rate decisions are assumed to be contemporaneously affected by oil price changes and exchange rate fluctuations. Oil price volatility can be seen as a possible threat to the inflation target, and it can be seen from the statements of the Central Bank that it includes oil prices when formulating its policy rate decisions. The restrictions on the policy rates can then be represented as:
𝛾41𝑒𝑜 + 𝛾52𝑒𝑒+ 𝛾54𝑒𝑒+ 𝛾55𝑒𝑜 = 𝑢𝑟 (16) This leaves us with the following structural model:
( 1 0 0 0 𝛾21 1 0 0 𝛾31 𝛾32 1 0 𝛾41 𝛾42 𝛾43 1 ) ( 𝑒𝑜 𝑒𝑒 𝑒𝑝 𝑒𝑟 ) = ( 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 ) ( 𝑢𝑜 𝑢𝑒 𝑢𝑝 𝑢𝑟 ) (17)
Note that this is a recursive ordering. Following Stock and Watson (2001), a different identification scheme was also applied for robustness, but that did not affect the results.33
33 Figure D4 in appendix D shows the impulse response functions for a VECM with 8 lags and the
Cholesky ordering: 𝑜𝑡→ 𝑟𝑡→ 𝑒𝑡→ 𝑝𝑡. This assumes that the Central Bank does react within-the-month to
5
Results
Following the discussion in section 4, two different VECM were analysed. The first model adopted a linear specification of oil prices (𝑜𝑡 variable), while the second model replaced the linear oil price variable with a non-linear specification (𝑁𝑂𝑃𝐼𝑡 variable).
5.1 Linear model
Figure 5 reports impulse response functions for real exchange rates, the CPI and policy rates to a one standard deviation shock to oil prices. The first thing one notices is that there is a significant increase in the consumer price index within-the-month. This indicates that oil price innovations do indeed increase inflation in Iceland. However, this hardly comes as a surprise since oil prices are directly included in the CPI. Secondly, in the aftermath of an oil price shock the real exchange rate of the króna depreciates after seven months. However, this decrease is only slightly significant for one month. Although there is a contemporaneous decrease, that decrease is not significant. The final and most interesting result from the oil price innovation is the reaction of the Central Bank. After an inflationary oil price shock, the Bank does not respond by tightening the monetary policy stance within-the-month, but rather deceases its policy rates significantly in the first four months. As a matter of fact, there is not a significant increase in the policy rate until after seven months.
