The effect of U.S. quantitative easing on inflation rates in EMEs

(1)

The Effect of U.S. Quantitative Easing on

Inflation Rates in EMEs

Pim Welling 10297529 Universiteit van Amsterdam

Bachelor Thesis Supervisor: Oana Furtuna

wellingpim@gmail.com

February 2, 2016

Abstract

Quantitative Easing has been a hot topic ever since the 2007-09 financial crisis, being a major fixture in US monetary policy. However, some of its effects tend to spill over to other countries, including the so-called emerging market economies. Through OLS and Dynamic OLS regressions, the effect of US Quantitative Easing on inflation rates in five emerging market economies, namely Brazil, Indonesia, India, South Africa, and Turkey, is estimated, resulting in no significant results as to whether the US Quantitative Easing program has had effect on the inflation rates of the said countries.

Keywords: quantitative easing, emerging market economies, unconventional monetary policy, inflation rate, spill over.

(2)

Statement of Originality

This document is written by Pim Welling, who declares to

take full responsibility for the contents of this document.

I declare that the text and the work presented in this

document is original and that no sources other than those

mentioned in the text and its references have been used in

creating it.

The Faculty of Economics and Business is responsible solely

for the supervision of completion of the work, not for the

contents.

(3)

1 Introduction

Quantitative Easing (QE) has been a hot topic during the 2007-09 financial crisis, being a major fixture of US monetary policy ever since. In short, QE can be defined as large-scale asset purchases (LSAPs), where central banks expand their reserves beyond the level needed to maintain its policy rate target (Mishkin, Matthews & Giuliodori, 2013, p.336). While QE is primarily used for lowering interest rates and promoting economic growth (Amadeo, 2013, p.1), some of the effects tend to spill over to other countries, including Emerging Market Economies (EMEs) (Chinn, 2013, p.1).

QE was however intended as a temporary policy, hence, on June 19, 2013, Chairman Bernanke announced the tapering of QE, stating that “…it would be appropriate to moderate the monthly pace of purchases…” (Bernanke, 2013, p.5). The actual tapering did occur in December that year (Amadeo, 2013, p.2). Initially, the market reacted heavily on this initial tapering announcement, which begs the question: what are the effects of US QE, and specifically, what are the effects of US QE on inflation rates in EMEs?

Extensive research has been done on the possible spill over effects of US QE, mostly focussing on capital flows, but considerably less information can be found about the spill over effects of US QE on inflation rates. To further research and determine potential effects, a econometric model is used on the so-called ‘fragile five’, five emerging market economies severely dependent on foreign investment, namely, Brazil, India, Indonesia, South-Africa and Turkey (Thomas, 2014, p.1). By regressing the inflation rate in these countries on multiple variables including US QE, potential effects can be determined. Data is collected OECD.Stat. and the Federal Reserve Bank of St. Louis, using datasets from December 31, 2006 to

December 31, 2014; periods during and after US QE.

This research could be helpful for the further use of unconventional monetary policy, specifically looking into its potential spill over effects. Policymakers get a clearer view on what effects monetary easing has on not only the countries implementing it, but also on developing countries dealing with an altered economic environment because of it.

The rest of this paper is organized as follows: in section two a literature studies is presented, with methodology briefly discussed in section three. Section four will,

(4)

2 Literature review

The phrase ‘Quantitative Easing’ was introduced to signal a shift in focus towards targeting quantity variables and was first coined during the real estate bubble and deflationary pressure Japan faced in the 1990s (Joyce et al., 2012, p.4).

However, the very first instance of QE has been reported as early as in 1932, when the American Federal Reserve (Fed), with congressional support, opted to buy $1 billion dollars worth of Treasury securities to stimulate the economy. US QE further continued during 1933-1936, eventually shifting to the outright purchasing of gold, whereby the US Treasury used the Gold Reserve Act (1934)1 to increase the value of the reserves of the Federal Reserve Banks (Anderson, 2010, p.1).

QE is known as an ‘unconventional monetary policy’. Central banks typically conduct policy embodying open market operations (OMOs); by either buying or selling securities, central banks influence the amount of reserves being held in the banking system, influencing short-term nominal interest rates (such as the overnight federal funds rate in the US) (Joyce et al., 2012, p.4; Bernanke, Reinhart & Sack, 2004, p.1).

Because conventional policy is predominately focused on maintaining low and stable inflation (Joyce et al., 2012, p.1), the change in reserves is viewed as merely a by-product of the policy, affecting short-term interest rates being the primary goal (Joyce et al., 2012, p.4). However, given that one wants to lower the short-term interest rate, what if it nears the so-called ‘zero lower bound’?, where the interest rate cannot be lowered further?

This situation is popularly known as the ‘liquidity trap’, which hinges on the school of thought that the nominal interest rate i is what matters most for economic policy and is, as a tool, unable to stimulate the economy anymore (Ito, 2008, p.1). In a recession, policymakers may even lower the real interest rate r into negative values (using the Fisher equation, r = i –

π, π being the level of inflation).

1

The Gold Reserve Act was passed on January 30, 1934 and raised the value of gold from $20.67 to $35 per ounce. On August 28, 1933 President Roosevelt called all outstanding domestic gold into the Federal Reserve Banks; this was before the gold revaluation. After the 1934 Gold Reserve Act revaluation the US Treasury started to purchase nearly $4 billion of gold, greatly increasing the monetary base (i.e. reserves held by the banking system) (Anderson, 2010, p.1).

(5)

However, once the nominal rate i reaches zero, it cannot be lowered further2, leaving the real interest rate at – πe (Ito, 2008, p.2).

Because bonds and money have now become close substitutes of each other due to the low interest level, people can simply choose to hold their money, and the financial injections from the central bank thus no longer comply. This is when further conventional monetary policy becomes ineffective (Fawley & Neely, 2013, pp.3-4; Ito, 2008, p.2).

By using QE, this liquidity trap is bypassed, and reserves and interest rates can again be influenced (Ito, 2008, p.4).

First and foremost, the aim of quantitative easing is to inject money into the economy to revive nominal spending. The central bank uses its reserves directly as a tool of monetary policy, purchasing assets and increasing the banking system reserves (Benford et al., 2009, p.2; McLeay, Radia & Thomas, 2014, p.24). The central bank thereby predominately buys from non-banks and financial institutions, whereby banks can function as intermediaries. To pay for these assets the central bank creates reserves, which was used to be done by “printing money”, but nowadays is done electronically (Bank of England, 2013, p.1). The previous holder of the asset can either spend the money on goods and services, which immediately affects the overall spending, or can buy different assets, which tends to up the asset prices. Higher asset prices will evidently lower yields, which in effect makes it less costly for consumers to lend. Due to the increased bank reserves, the amount of loans a bank can write increases, boosting the economy (Bank of England, 2013, p.1; Benford et al., 2009, p.2).

Yet the effectiveness of the asset programmes such as QE in supporting the economy has not been universally accepted, as Chen et al. (2012) notes that critics have voiced their concern before initial implementation.

2.1 US QE and the 2007-08 global financial crisis

As the collapse of the sub-prime mortgage market and the real estate bubble ushered in the financial crisis of 2007-09, US policy makers opted for the use of unconventional monetary tools to combat the crisis head-on. Where the European Central Bank (ECB) made use of

2_{There have been instances that the zero lower bound has been surpassed, such as in Denmark (Joyce et al.,}

(6)

increased bank lending, the Fed went ahead with QE3 (Morgan, 2011, p.3). Early on, these asset purchases focused on mortgage backed securities and US Treasury bonds, whereby the purchases of Treasury bonds were deemed primarily to stimulate economic activity;

supporting investment and lending during the crisis. Subsequently it evolved to support the post-crisis economic recovery in growth and employment (Fratzscher et al., 2013, pp.7-9; Lim, Mohapatra & Stocker, 2014, p.5).

In the US, the asset purchases have covered a range of different types of assets, such as commercial paper and asset-backed securities (Benford et al., 2009, p.2). The US policy makers first aimed at the lowering of bond yields and spreads, not targeting the reserves per se. However, as a consequence, the LSAPs have increased the Feds reserves greatly; these assets may stay on the Feds balance sheets for numerous more years to come (Gagnon et al., 2010, p.1).

On November 25, 2008, the Fed made the first announcement of purchases worth of $100 billion in government-sponsored enterprise debt (GSE) and $500 billion worth of Agency Mortgage-Backed-Securities (AMBS). Followed by an other announcement on March 18, 2009 for another purchase of $100 billion in GSE, $750 billion in AMBS and $300 billion in long-term US Treasury securities, these two asset purchasing programs combined are

commonly called the first round of quantitative easing, or QE1. Although the program was focused on the economy as a whole, its main priority lied within the housing markets, with GSE debt and AMBS taking up more than 80% of the assets purchased by the Fed in its first round of QE (Fawley & Neely, 2013, p.60).

In October 2010, with the continuing economic weakness and still facing the zero lower bound, the Fed announced another round of QE: QE2. Now on a smaller scale and predominately focussed on purchasing long-term government bonds, involving $600 billion long-term US Treasuries (Fratzscher et al., 2013, p.2). This was again with intent to stimulate the US economy (specifically employment), but now by lowering yields and pushing up asset prices in the higher risk markets (D’Amico et al., 2012, p.2).

QE1 worked mainly through portfolio rebalancing across multiple countries, while QE2 worked through portfolio rebalancing across different sort of assets and across countries (Fratzscher, 2013, p.5).

3_{The Fed itself did not characterize these asset purchases as QE, and instead referring to them as LSAPs}

(7)

The QE2 program has been controversial, with critics claiming the costs of the policy are large while the benefits are small (Swanson et al., 2011, p.3).

In September 2011, the Fed initiated another slightly adjusted version of QE2, a maturity extension program popularly known as ‘Operation Twist’; the twist being the way the purchases were being executed. The term came from a similar program by the Kennedy administration in 1961-64, whereby short-term Treasury was sold and long-term Treasury debt was purchased. The ‘Operation Twist’ program aimed at modifying the term structure of the interest rates and showed a commitment to extend the maturity of securities held on the Feds balance sheet. As such, the total quantity of assets on the balance sheet did not change. (Chen et al., 2012, p.231; Fratzscher et al., 2013, p.4; Tillman, 2015, p.2; Amadeo, 2013, p.1).

Bernanke announced that the Fed would maintain its expansive policy until employment further improved (Amadeo, 2013, p.1); and so, September 12, 2012 was the starting point for the QE3 program, which primarily focussed on purchasing AMBS (Gertler & Karadi, 2013, pp.6-7).

On December 18, 2013, the Fed announced it would begin tapering its purchases, as unemployment rates were down, GDP showed growth rates around 2-3% and inflation remained somewhat stable, not exceeding 2% (Amadeo, 2013, p.3). In total, the Fed engaged in three separate QE rounds; when the Fed had began its initial downgrading of QE in January 2014, its balance sheet had increased more than twofold, to $4 trillion dollars

(Lim, Mohapatra & Stocker, 2014, p.2).

2.2 US QE and Emerging Market Economies

The US QE program was initially meant to be an expansionary policy for the US economy, but has had profound effects on developing countries4. Given the fact that during US QE interest rates in the US neared the zero lower bound and the Fed upped their LSAPs, financial institutions began looking for alternative sources of investment. EMEs had enjoyed high growth rates and a relatively stable political environment in the last ten years, thus appearing to be a formidable alternative (Lim, Mohapatra & Stocker, 2014, p.2).

4_{Morgan (2011) estimates that about 40% of the increase of US reserves spilled over in the form of private}

(8)

The idea that policies like QE generally have substantial cross-border impact has been supported by numerous recent studies, notably Chinn (2013), Tillman (2015) and Fratzscher

et al. (2012), for example stating that monetary policy changes in larger countries cause big

swings in commodity prices for other countries.

Researchers can clearly see how the QE announcements affect asset prices, such as bond yields and exchange rates, because asset prices react swiftly to new information (Neely, 2014, p.1). Chinn (2013) further states that because the standalone effects of the

announcements of QE and its tapering5, it would be difficult to denounce any spill over effects QE has had in itself6.

Literature has cited numerous channels through which QE spill over effects affect EMEs, and specifically its capital flows;

Portfolio-balance channel: with QE, long-term assets, such as mortgage-backed securities and

long-term bonds, are being purchased. This increased purchasing reduces the supply of the available stock of privately-held risk assets to the private sector, and thereby increasing the private holdings of short-term risk-free bank reserves. For investors to work with this, the initial return of the asset being purchased must fall, i.e. the yield must be lower (Gagnon, 2010, p.3). On the other hand, investors can turn to more riskier assets because of the unmet risk appetite. This will lower risk-premiums, ups their prices and lowers their yields (Lavigne

et al., 2014, p.3; Lim, Mohapatra & Stocker, 2014, p.6).

So in this regard, it is expected that there will be a rise in private demand for longer duration and developing country assets, which leads to a rebalancing of investor portfolios (Lim, Mohapatra & Stocker, 2014, p.6).

Liquidity channel: another key transmission channel for QE affecting capital flows. Due to the

programs asset purchases, private banks can denote increased reserves on their balance sheets. These reserves can be traded on the secondary market; more easily than the long-term

securities, leading to a decline in the liquidity premium. This helps previously

5_{The initial talking of tapering in 2013 alone have caused disruptive effects on the capital inflows to EMEs}

(Lavigne et al., 2014, p.2; Chinn, 2013, p.2).

6_{Fratzscher (2013) notes that the US QE announcements had comparative less effect than the actual QE}

(9)

constrained banks to give out more credit to investors, further enabling banks to extend credit to investors (Lim, Mohapatra & Stocker, 2014, pp.6-8).

Signalling channel: US QE can be read as a commitment by the Fed, because it transmits

information about the future monetary policy, namely keeping interest rates low even after the economy recovers. These may influence the bond yields negatively, which in turn may lead to a larger span between interest rates in the US and the EMEs. This increased spread leads to an increased trade and capital inflow for EMEs (Lim, Mohapatra & Stocker, 2014, pp.6-8; Tillman, 2015, p.2).

Exchange rate channel: The changes in portfolios as described earlier could cause a

depreciation of the US dollar, which in turn would lead to a decrease in US foreign demand, predominately in relative terms with the home-produced goods. This would have a negative effect on the EMEs export mechanism (Lavigne et al., 2014, p.3), thereby influencing the inflation rate negatively, working against the earlier mentioned channels.

Trade-flow channel: due to the increased money supply, US QE could directly lead to

increased EME exports, which could offsets the negative export and inflation effects of the earlier described exchange rate channel (Lavigne et al., 2014, p.3).

As can be seen, given the magnitude of these channels it is difficult to estimate a precise effect of US QE on EME inflation rates. These channels work simultaneously with and against each other with respect to inflation. However, Lavigne et al. (2014) have summarized that US QE has in al likeliness increased the capital inflows to EMEs, and put upward

pressure on asset prices. These capital inflows were however also supported by strong economic fundamentals the EMEs have developed.

MacDonald (2015) further states that US QE has had large currency appreciations, decreases in long-term local currency sovereign yields, and increases in equity markets across EMEs. The degree to which individual EMEs were affected, showed to be heterogeneous, which could be attributed to the degree of capital market frictions between EMEs and the US (MacDonald, 2015 p.21). EME policy-makers may also have tried to shield themselves from the spill over effects, for example by interventions in foreign exchange markets and

(10)

introducing capital controls7. Countries with better institutions and which are more active in monetary policy have been effected less by QE (Fratzscher, 2013, p.7).

All in all, literature suggests that the impact of US QE on EMEs overall could likely be characterised as positive, predominately due to positive trade and confidence effects which stem from a stronger economic activity in the US (Morgan, 2011, p.1; Lavigne, et al., 2014, p.1).

Foreign-policy-makers still have voiced their critiques and concern about US QE, fearing policy spill overs through an uncontrolled increase of cross-border capital flows (Lim,

Mohapatra & Stocker, 2014, p.5). Thereby stating that US QE has in fact created an excessive global liquidity which was caused by the increased capital inflows to EMEs since 2009. Policy makers suggest that these inflows have caused appreciation pressures on the EMEs currencies, financial imbalances8, asset-price bubbles and an over-heating of the emerging markets domestic economies. November 18th, 2010, China’s vice minister of finance, Zhu Guangyao, stated "As a major reserve currency issuer, for the US to launch a second round of quantitative easing at this time, we feel that it did not recognise its responsibility to stabilize global markets and did not think about the impact of excessive liquidity on emerging

markets." (Morgan, 2011, p.3).

Some EME policy-makers have suggested that the LSAPs have promoted unnecessary risk taking and increased larger-than-normal capital inflows to EMEs. Thereby also noting that the upward pressure on asset prices and exchange rates has deteriorated EMEs

competiveness, also voicing their concern that once the conventional monetary policy once again sets in, there is a risk of disruptive capital withdrawals from EMEs (Lavigne et al., 2014, p.1).

Initial concerns of uncontrollable capital inflows have since shifted to the fear of reversal of capital flows back into mature economies. Since the 2013 tapering announcement

policymakers fear they would struggle the consequences of a ”sudden stop of inflows or even a reversal of flows” (Tillman, 2015, p.2). The prospect of tapering of QE by advanced

7

Brazil, Indonesia, and South Korea among others (MacDonald, 2015, p.1).

8_{Recipient countries of QE do tend to have increased local leverage, meaning that due to the rising asset prices}

and appreciated exchange rate, borrowers can seem to have a greater equity than they truly have (Rajan, 2014, pp.6-7).

(11)

economies, including the US, has been associated with nominal exchange-rate depreciations, stock market contractions (MacDonald, 2015, p.1), and other problematic events in EMEs due to the altering of earlier speculative positions by investors (Lim, Mohapatra & Stocker, 2014, p.6; Aizenman et al., 2014, p.2).

As stated earlier, when then Chairman of the Fed, Ben Bernanke, did on 22 May 2013 announced QE tapering for the first time, policy expectations changed immediately. Risk tolerance likely decreased and caused a reassessment of the potential returns of investing in EMEs. When the long-term rates promptly rose, private capital inflows to EMEs decreased, and market volatility increased, sparking significant capital outflows from numerous EMEs (Lim, Mohapatra & Stocker, 2014, p.6).

However, Lavigne et al. (2014) states that after the initial announcement of QE tapering, capital flows of EMEs became more individualized. Capital inflows of EMEs were likely correlated to the country specific macroeconomic fundamentals, which in some regard reflected the policies the countries went with after the financial crisis of 2007-09.

3 Methodology

As stated earlier by Joyce et al. (2012), the main aim of monetary policy is to achieve low and stable inflation. When experiencing increased capital inflows as an EME, one could argue that this increase has a notable effect on inflation. The tapering of these capital flows may once again influence inflation, asking for monetary policy action. Literature states that US QE spill overs indeed have caused increased capital flows to EMEs (Morgan, 2011, p.1).

By using numerous macroeconomic variables it can be estimated to what amount the US QE program has affected the inflation rate in these ‘fragile five’; Brazil, Indonesia, India, South Africa, and Turkey.

Inflation is influenced by QE predominantly through asset prices and nominal spending (Benford et al., 2009, p.6). This goes by several transmission mechanisms such as the earlier mentioned balance channel and the liquidity channel. As for the portfolio-balance channel, the result of US QE has been that indeed, borrowing costs were lowered for corporations and households, which stimulated spending (Joyce, Tong & Woods, 2011, p.2). This channel would therefore affect the EMEs inflation rates positively, due to the increase in spending. The same would go for the signalling and liquidity channel, given that they both stimulate capital inflows to EMEs, which as mentioned earlier, could be somewhat interpreted as increased levels of inflation.

(12)

The exchange rate channel affects the export negatively, and would impact inflation negatively, whereas the trade-flow channel could offset these negative effects.

Benford et al. (2009) argues that increased nominal spending and asset prices would increase inflation, but that by signalling from the central bank market participants may think that the lower rates assume that its trying to keep its inflation target. This would eventually have a negative effect on the inflation9 (Joyce et al., 2012, p.2). However, it is all dependent of different strengths and prevalence of different transmission channels, making it very difficult to estimate an empirical model with confidence.

As for the models used, the choices of variables, the decision to include or exclude certain parts, was all done with a combination of clear intuition and the use of existing literature.

The hypothesis then becomes, following the findings of Lavigne et al. (2014): the effect of US QE is positively correlated with the inflation rate of EMEs.

So, to estimate the effects and impact of the variables on the inflation rate, a multiple linear OLS regression model is used with dependent and independent variables; assuming the independent variables will have some effect on the dependent inflation.

A second regression is performed, using the so-called dynamic OLS model, which accounts for specific time lags and dummy variables.

3.1 OLS regression

𝜋𝜋 = 𝛽𝛽0 + 𝛽𝛽1∗ 𝐼𝐼𝐼𝐼 + 𝛽𝛽2∗ 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝐼𝐼 + 𝛽𝛽3∗ 𝐼𝐼𝑅𝑅𝑊𝑊 + 𝛽𝛽4∗ 𝐸𝐸𝐸𝐸𝐼𝐼 + 𝛽𝛽5∗ 𝑄𝑄𝐸𝐸 + 𝜀𝜀𝑖𝑖 As can be seen, the model has one dependent variable and five independent ones.

𝐼𝐼𝑈𝑈𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑈𝑈 𝑟𝑟𝐼𝐼𝐼𝐼𝑈𝑈 𝜋𝜋

As for the dependent variable, inflation, one has to find a good indicator to measure this in a specific country. Literature suggests that the Consumer Price Index (CPI), which is a fixed wage index for the cost of living (Bryan & Cecchetti, 1993, p.2), is the go-to statistical

9_{Because the inflation is measured in countries only affected by the spill over effects of QE, and not in}

countries implementing QE themselves, the signalling from the central bank to keep the inflation stable and low has not been included in both empirical models.

(13)

indicator for inflation rate (Falnita & Sipos, 2007, p.2). By looking at the way this CPI evolves over time, a educated guess can be made to what the inflation rate upholds.

Hence,

100 ∗ �𝐶𝐶𝐶𝐶𝐼𝐼𝑡𝑡_{𝐶𝐶𝐶𝐶𝐼𝐼}− 𝐶𝐶𝐶𝐶𝐼𝐼𝑡𝑡−1

𝑡𝑡−1 � = 𝐼𝐼𝑈𝑈𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑈𝑈 𝑟𝑟𝐼𝐼𝐼𝐼𝑈𝑈 𝜋𝜋

The data is retrieved from OECD.Stat.10 Monthly observations were used for a total of 97 observations, from 31-12-2006 up until 31-12-2014. These were periods before, during, and after the US QE program ran.

The inflation rate data is found as the growth of consumer prices of the earlier year, thereby using the earlier mentioned inflation rate formula.

(𝐼𝐼𝑈𝑈𝐼𝐼𝐼𝐼) 𝐼𝐼𝑈𝑈𝐼𝐼𝑈𝑈𝑟𝑟𝑈𝑈𝐼𝐼𝐼𝐼 𝑟𝑟𝐼𝐼𝐼𝐼𝑈𝑈 𝐼𝐼𝐼𝐼

To speak in general terms, once the interest rates are lowered, people can borrow at a lower rate. Because of the lower rate, more people can afford to lend and in fact will, increasing the nominal spending. Hence, the economy grows, and so will the inflation. Once the interest rates are being increased the opposite will happen and we would speak of a deflation effect.

Noted earlier, the Fisher equation involves inflation and the real and nominal short-term interest rates.

The approximation of the formula (ex post),

𝑟𝑟 ≈ 𝐼𝐼 − 𝜋𝜋

The formula states that the real interest rate r is approximately the nominal interest rate i minus the rate of inflation π (Horn, 2008, p.3). Once the real interest rates rise, the inflation would shrink according to this equation.

For the model, the intermediate interest rates were taken, as Pericoli (2014) has shown that these rates have been affected by QE spill overs. Data has been retrieved from

OECD.Stat. and uses 97 observations ranging from 31-12-2006 to 31-12-2014.

10

(14)

𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝐼𝐼𝐼𝐼𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝐼𝐼 𝑟𝑟𝐼𝐼𝐼𝐼𝑈𝑈 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝐼𝐼

The unemployment rate influences the inflation rate. One could simply argue that because more people work for pay, more people spend more money. This in all increases the money in the economy, leading to inflation. To look at the unemployment rate, monthly data will be used to determine the effects on inflation, namely 97 observations from 2006 to 31-12-2014.

The inflation – unemployment trade-off has been described in literature by the Philips curve. This short-term trade-off is closely related to the concept non-accelerating inflation rate of unemployment (NAIRU), this is the rate of unemployment where the inflation has no tendency to change. Simply put, the Philips curve theory states that increased unemployment (above NAIRU) corresponds with a lower inflation cq. deflation.

However, many economists have questioned its validity, stating that the curve should not be used as a guide or indicator for monetary policy conduct (Mishkin, Matthews & Giuliodori, 2013, p.370).

The unemployment rate data comes from OECD.Stat., however, not all five EMEs have monthly unemployment rate data during the period of interest. Therefore, in the regression some data is missing, which could lead to a skewed view in some of the results.

𝐹𝐹𝐼𝐼𝐹𝐹. 1 𝐶𝐶ℎ𝐼𝐼𝐼𝐼𝐼𝐼𝑈𝑈𝐼𝐼 𝑐𝑐𝑐𝑐𝑟𝑟𝑐𝑐𝑈𝑈, 𝐶𝐶ℎ𝐼𝐼𝐼𝐼𝐼𝐼𝑈𝑈𝐼𝐼 (1958)

𝐼𝐼𝐼𝐼𝑅𝑅 𝑈𝑈𝐼𝐼𝐼𝐼𝑈𝑈𝑟𝑟𝐼𝐼𝐼𝐼𝐼𝐼 𝑈𝑈𝑟𝑟𝐼𝐼𝑐𝑐𝑈𝑈𝐼𝐼 𝐼𝐼𝑅𝑅𝑊𝑊

Raw material prices have a profound effect on inflation. This mainly because of the fact that raw materials play an important role in many different layers of the economy, being part of an enormous amount of products and services. Experiencing higher raw material prices would result in higher input costs, leading to higher end product prices and services.

(15)

For the 𝐼𝐼𝑅𝑅𝑊𝑊 variable international crude oil prices are taken as an indicator (using WTI crude oil as benchmark price). Crude oil prices for instance affect products which have plastic in it, affect fuel costs, and by doing so affect the economy as a whole. An increase in oil prices would lead to a higher price level, which is shown to correlate with an increase in inflation (Mishkin, Matthews & Giuliodori, 2013, pp.10-11).

Monthly data is used from Federal Reserve Bank of St. Louis11, with 97 observations ranging from 31-12-2006 to 31-12-2014. This data is independent of the EME in question, the same international prices (in dollars per barrel) are used for every country.

(𝐼𝐼𝑈𝑈𝐼𝐼𝐼𝐼) 𝐸𝐸𝐸𝐸𝑐𝑐ℎ𝐼𝐼𝑈𝑈𝐹𝐹𝑈𝑈 𝑟𝑟𝐼𝐼𝐼𝐼𝑈𝑈 𝐸𝐸𝐸𝐸𝐼𝐼

In many countries, adjusting the real exchange rate has been a way to influence inflation. Literature suggests that Brazil, Indonesia, India, South Africa, and Turkey all have a floating exchange rate regime, which makes the effect of exchange rate differences on inflation and vice versa observable.

Edwards (2006) argues that the effect of the exchange rates on inflation is dependent on as to how much does ‘pass-through’ from exchange rates to domestic prices. Which in turn would correlate with the inflation rate.

The real exchange rate differs from the nominal exchange rate in the sense that it is adjusted for the relative prices between countries (Pilbeam, 2013, p.11).

In formula form,

𝑆𝑆𝑟𝑟 = 𝑆𝑆𝐶𝐶 ∗ 𝐶𝐶

Where 𝑆𝑆_𝑟𝑟 is the real exchange rate, 𝑆𝑆 is the nominal exchange rate, and 𝑃𝑃∗

𝑃𝑃 is the foreign price level divided by the domestic price level.

Pilbeam (2013) argues that if the real exchange rate appreciates, as such that the said currency is relatively worth less, inflation will rise. Due to the fact that imports will become more expensive, the price level of imported goods will increase. On the other hand, because the export is growing due to the low export prices, the economy as a whole is growing. This would mean that the exchange rate would have a positive correlation with the inflation rate.

11

(16)

Once again, data has been retrieved from OECD.Stat., 97 observations from 31-12-2006 to 31-12-2014.

𝑈𝑈𝑆𝑆 𝑄𝑄𝑐𝑐𝐼𝐼𝑈𝑈𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑐𝑐𝑈𝑈 𝐸𝐸𝐼𝐼𝐼𝐼𝐼𝐼𝑈𝑈𝐹𝐹 𝑄𝑄𝐸𝐸

To include US QE in the model, a specific indicator is necessary to simulate the variable. As described earlier, the Fed purchased large amounts of US Treasuries and Mortgage-Backed Securities (MBS) during its QE program. Therefore to include the amount of US Treasuries and MBS held by the Fed recreates the economical effect of US QE on these EMEs.

The data on these assets held is retrieved from the databank of the Federal Reserve Bank of St. Louis, monthly data which shows these assets held by the Fed. This data is independent of the countries in question, as this data describes information in the US (97 observations from 31-12-2006 to 31-12-2014).

3.2 Dynamic OLS

Inflation is known to take its time to fully respond to changes in monetary policy. Batini & Nelson (2001) state that it at least takes a year for monetary policy to see a full-on effect on the inflation rate.

This would mean that because of this apparent lag, the effect of inflation of (at least) one year ago must be accounted for. Therefore, three new variables are added to include this lag-effect. Hence, the effect of e.g. inflation of twelve months ago on the inflation at the current date t. Because of the inclusion of the phrase ‘at least’ in the work of Batini & Nelson, lag variables for twelve, thirteen and fourteen months are added to the model.

This step calls for a dynamic model which requires to de-string the data and make the regression a dynamic one. This step also makes room to include two extra time dummy variables, which account for the dates for which, respectively, the Fed announced the first US QE purchases and the Fed announced its initial tapering of US QE.

The use of a dynamic model is more logical given the apparent lags, and therefore more important for the conclusions that can be drawn from this paper.

The model changes, hence,

𝜋𝜋 = 𝛽𝛽0+ 𝛽𝛽1∗ 𝐼𝐼𝐼𝐼 + 𝛽𝛽2∗ 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝐼𝐼 + 𝛽𝛽3∗ 𝐼𝐼𝑅𝑅𝑊𝑊 + 𝛽𝛽4∗ 𝐸𝐸𝐸𝐸𝐼𝐼 + 𝛽𝛽5∗ 𝑄𝑄𝐸𝐸 + 𝛽𝛽6∗ 𝐿𝐿𝐼𝐼𝐹𝐹1 + 𝛽𝛽7 ∗ 𝐿𝐿𝐼𝐼𝐹𝐹2 + 𝛽𝛽8 ∗ 𝐿𝐿𝐼𝐼𝐹𝐹3 + 𝛽𝛽9∗ 𝐹𝐹𝐼𝐼𝑟𝑟𝐼𝐼𝐼𝐼. 𝑑𝑑 + 𝛽𝛽10∗ 𝑇𝑇𝐼𝐼𝑈𝑈𝑈𝑈𝑟𝑟𝐼𝐼𝑈𝑈𝐹𝐹. 𝑑𝑑 + 𝜑𝜑𝑖𝑖

(17)

The model stays the same as the OLS regression model, but now includes five new variables.

𝐼𝐼𝑈𝑈𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑈𝑈 𝐿𝐿𝐼𝐼𝐹𝐹 1, 2 & 3

Three lag variables that are included for inflation. Mentioned earlier, the three lag variables account for lags of twelve to fourteen months.

𝐹𝐹𝐼𝐼𝑟𝑟𝐼𝐼𝐼𝐼 𝐼𝐼𝑈𝑈𝐼𝐼𝑐𝑐𝑈𝑈𝑐𝑐𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝐼𝐼 𝑄𝑄𝐸𝐸 𝐹𝐹𝐼𝐼𝑟𝑟𝐼𝐼𝐼𝐼. 𝑑𝑑

A variable which accounts for the announcement for the first QE purchases of the Fed. This is a time dummy variable for the date November 25, 2008. However, because the data is

monthly, the dummy is for November 2008.

𝑇𝑇𝐼𝐼𝑈𝑈𝑈𝑈𝑟𝑟𝐼𝐼𝑈𝑈𝐹𝐹 𝐼𝐼𝑈𝑈𝐼𝐼𝑐𝑐𝑈𝑈𝑐𝑐𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝐼𝐼 𝑇𝑇𝐼𝐼𝑈𝑈𝑈𝑈𝑟𝑟𝐼𝐼𝑈𝑈𝐹𝐹. 𝑑𝑑

A time dummy variable for the date December 18, 2013, the date where the Fed announced its initial tapering of US QE. Once again, because of the monthly data the dummy is for

December 2013.

3.2.1 Durbin-Watson statistic

When using the regression with all its variables, a Durbin-Watson test statistic can be used to test for autocorrelation, which is when error terms of any of the observations is correlated with one another. This seems to be primarily the case in time-series, where data is gathered in different time periods repeatedly (“Serial Correlation, the Durbin-Watson Statistic, and the Cochrane-Orcutt Procedure”, 2006, p.1).

First off, the Durbin-Watson statistic must be computed, which then can be interpreted with critical values, checking if there might be some form of autocorrelation. For all the countries there appeared to be some positive autocorrelation, falling between either positive serial correlation or indeterminate (no decision).

Positive autocorrelation Indeterminate No autocorrelation Indeterminate Negative autocorrelation

(18)

This was then corrected with the Cochrane-Orcutt procedure, which is stated in the last columns of the dynamic OLS regression output. As can be seen in figure two, the newly transformed Durbin-Watson-statistics are superior to the original values in terms of lower levels of autocorrelation.

4 Results & Conclusion

As for the results, the regression output for Turkey in OLS and Dynamic OLS, respectively, is shown below, the output for the other countries can be found in the appendix.

The OLS regression shows that in the full model, the independent variable QE does seem to have an effect, being significant in all countries except for Brazil12. The result does not correspond to the earlier findings of Barosso, Pereira da Silva & Soares Sales (2015), which concludes that US QE did indeed have a positive effect on capital inflows in Brazil.

12

* indicates significant at the 5% level.

Country DW-statistic (original) DW-statistic (transformed) (dL; dU)

Turkey 1.042987 1.800441 (1.262; 1.773)

Brazil 0.3739903 1.323603 (1.072; 1,817)

India 1.484186 2.23705 (0.507; 2.097)

Indonesia 0.9483232 1.683616 (1.162; 1792)

South Africa 0.5643199 1.556873 (1.262; 1.773)

Fig. 2 Durbin-Watson statistics & correction

Turkey

Dependent variable: Inflation rate Method: OLS

Included observations: 29 monthly observations

Regressor (1) (2) (3) (4) (5)

QE 2.27e-07 6.22e-07* 3.60e-07 -3.51e-07 -9.48e-07*

IR 0.2625235* 0.2547061* 0.2098865* 0.2105408* EXR 0.992649 2.275562 4.394729* UnemR -0.3951542* -.2527684 RAW 0.03207* Constant 8.70948 4.730051 3.694626 7.737543 1.127141 Root MSE 1.8312 1.5523 1.5563 1.4879 1.4244 Adj. R2 0.0101 0.2886 0.285 0.3465 0.401

(19)

This deviance might occur because of the workings of the transmission mechanisms in Brazil. However, as mentioned earlier, inflation is such a broad subject in economics that it is frankly almost undoable to account for all variables and factors, so not being significant in Brazil could have a wide array of underlying reasons.

The coefficients for QE do all seem to be very small, being either negative or positive, indicating that the effects of the Feds’ QE program seem to be very limited. Apart from the amount of variables included in the regression, this limited effect might also be attributed to the manner in which the variable QE is estimated, namely by the changes in the amount of MBS and US Treasuries bought by the Fed. The moment that these assets are bought up, doesn’t necessary match the moment when for example the commercial banks can once again spend these reserves. There might be a time lag there where the model has not accounted for.

In column (5) we see the dynamic model with all its variables included, and it can be seen that, looking at the output of all the countries, the variable QE is significant in less countries than in the OLS regression. This can be due to the fact that this time there are more variables accounted for. Now, Turkey and South Africa do not seem to experience significant effects on

Turkey

Dependent variable: Inflation rate Method: Dynamic OLS, robust

Regressor (1) (2) (3) (4) (5) (5), corc

QE -3.4e-07* 4.00e-07* 6.29e-07* -4.83e-09 -1.74e-07 -6.01e-07

IR 0.3324972* 0.3453794* 0.3297713* 0.314548* .2922729* EXR -0.8603884 0.2830544 0.9536025 2.603243 UnemR -0.3963641* -0.3288856* -.197218 RAW 0.0085783 0.023841* Lag1 -0.7989517* -0.763789* -0.7496488* -0.7279976* -0.7208108 -.5784877* Lag2 -0.0469505 -0.0023563 -0.0044439 -0.0127336 -0.0168182* -.0012166 Lag3 0.3203416 0.2401385 0.2272278 0.2998945* 0.2852385* .1181631 First.d 2.601343 0.6795837 0.84572 0.1436159 0.3047412 -.2294146 Tapering.d -1.753366 -1.502442 -1.45091 -1.402853 -1.460583 .0628061 Constant 13.15946 9.149328 10.01586 13.27137 11.20529 6.778391 Root MSE 1.4844 0.96749 0.96944 0.846 0.84669 .69894 Adj. R2 0.3927 0.742 0.96944 0.8027 0.8024 0.5263

(20)

their inflation rates. This deviation can once again be attributed to the quality of the model itself, which is simply too small to account for the many channels and variables that influence inflation, and does not have the proper adjustments to correct for all the transmission channels at play.

The coefficients appear to be even smaller, but still seem to alternate between negative and positive values, reaffirming the earlier stated conclusion that the effect of US QE on these ‘fragile five’ EMEs is very limited.

For India, the regression is somewhat different in the sense that because of missing data in the unemployment variable, the dummy variable 0 and the 0’s of the non-data of unemployment began to correlate, resulting in multicollinearity. Therefore, more regressions are done, either dropping the dummy variables or the unemployment variable.

The last columns, (5), corc, are the same regressions as the ones stated in column (5), but this time corrected for autocorrelation. With the Cochrane–Orcutt procedure, new

coefficient estimates are found, resulting in even less significant effects for the QE variable. The size of the QE-coefficients does change, but do remain very small.

Conclusively, it can be said that the results of these regressions are far from

significant. From the regression output, OLS and Dynamic OLS, it can be seen that QE does not always have a significant impact on inflation rates, and more importantly, the effect fluctuates between positive and negative between different regressions and countries. This would mean that the hypothesis that QE did indeed have a positive effect on the inflation rates of Brazil, Indonesia, India, South Africa and Turkey, is rejected. However, this is not to be said if a more comprehensive model and statistical method is used. But for now, the results of this paper remain, sadly, inconclusive.

(21)

References

Aizenman, J., Binici, M., & Hutchison, M. M. (2014). The transmission of Federal Reserve

tapering news to emerging financial markets (No. w19980). National Bureau of

Economic Research.

Amadeo, K. (2013). What Is Quantitative Easing? Retrieved from http://useconomy. about. com/od/glossary/g/Quantitative-Easing

d’Amico, S., English, W., López-Salido, D., & Nelson, E. (2012). The Federal Reserve's Large-scale Asset Purchase Programmes: Rationale and Effects*.The Economic

Journal, 122(564), F415-F446.

Anderson, R. G. (2010). The First U.S. Quantitative Easing: The 1930s. St. Louis Fed: Economic Research. Retrieved from

https://research.stlouisfed.org/publications/es/10/ES1017.pdf

Benford, J., Berry, S., Nikolov, K., Young, C., & Robson, M. (2009). Quantitative easing. Bank of England Quarterly Bulletin, 49(2), 90-100.

Bernanke, B., Reinhart, V., & Sack, B. (2004). Monetary policy alternatives at the zero bound: An empirical assessment. Brookings papers on economic activity, 2004(2), 1-100.

Bryan, M. F., & Cecchetti, S. G. (1994). Measuring core inflation. In Monetary Policy (pp. 195-219). The University of Chicago Press.

Chen, H., Cúrdia, V., & Ferrero, A. (2012). The Macroeconomic Effects of Large-scale Asset Purchase Programmes*. The economic journal, 122(564), F289-F315.

Chinn, Menzie David, Global Spillovers and Domestic Monetary Policy (December 2013). BIS Working Paper No. 436. Available at SSRN: http://ssrn.com/abstract=2384451 Falnita, E., & Sipos, C. (2007). A multiple regression model for inflation rate in Romania in

the enlarged EU. Retrieved from https://mpra.ub.uni-muenchen.de/11473/

Fawley, B. W., & Neely, C. J. (2013). Four stories of quantitative easing. Federal Reserve

Bank of St. Louis Review, 95(January/February 2013).

Fratzscher. Marcel and Lo Duca, Marco and Straub, Roland, On the International Spillovers of US Quantitative Easing (March 2013). DIW Berlin Discussion Paper No. 1304. SSRN: http://ssrn.com/abstract=2276855 or http://dx.doi.org/10.2139/ssrn.2276855

(22)

Gagnon, Joseph and Raskin, Matthew and Remache, Julie and Sack, Brian P., Large-Scale Asset Purchases by the Federal Reserve: Did They Work? (March 1, 2010). FRB of New York Staff Report No. 441. Available at

SSRN: http://ssrn.com/abstract=1576596 or http://dx.doi.org/10.2139/ssrn.1576596 Gertler, M., & Karadi, P. (2013). Qe 1 vs. 2 vs. 3...: A framework for analyzing large-scale

asset purchases as a monetary policy tool. international Journal of central

Banking, 9(1), 5-53.

Giuliodori, M., Matthews, K. & Mishkin, F.S. (2012). The Economics of Money, Banking and Financial Markets: European edition. United Kingdom: Pearson Higher

Education.

Ito, H. (2008). Liquidity Trap. Retrieved from http://web.pdx.edu/~ito/Liquidity_trap_RE_-HI2.pdf

Joyce, M., Miles, D., Scott, A., & Vayanos, D. (2012). Quantitative Easing and

Unconventional Monetary Policy–an Introduction*. The Economic Journal,122(564), F271-F288.

Joyce, Michael and Tong, Matthew and Woods, Robert, The United Kingdoms’s Quantitative Easing Policy: Design, Operation and Impact (September 19, 2011). Bank of England Quarterly Bulletin, 2011. Available at SSRN: http://ssrn.com/abstract=1933696 Lavigne, R., Sarker, S., & Vasishtha, G. (2014). Spillover Effects of Quantitative Easing on

Emerging-Market Economies. Bank of Canada Review,2014(Autumn), 23-33.

Lim, Jamus Jerome and Mohapatra, Sanket and Stocker, Marc, Tinker, Taper, QE, Bye? The Effect of Quantitative Easing on Financial Flows to Developing Countries (March 1, 2014). World Bank Policy Research Working Paper No. 6820. Available at

SSRN: http://ssrn.com/abstract=2417518

McLeay, Michael and Radia, Amar and Thomas, Ryland, Money Creation in the Modern Economy (March 14, 2014). Bank of England Quarterly Bulletin 2014 Q1. Available at SSRN: http://ssrn.com/abstract=2416234

Morgan, Peter, Impact of US Quantitative Easing Policy on Emerging Asia (November 18, 2011). ADBI Working Paper No. 321. Available at

SSRN: http://ssrn.com/abstract=1961492 or http://dx.doi.org/10.2139/ssrn.1961492 Neely, C. J. (2014). Lessons from the taper tantrum. Economic Synopses,2014(2014-01-28).

(23)

Pericoli, M. (2014). Interest Rate Spillovers during Quantitative Easing. Bank of of Italy. Retrieved from

https://www.ecb.europa.eu/events/pdf/conferences/140908/pericoli.pdf?0a88aa6df59e a0b2e3260ed2162f286d

Pilbeam, K. (2006). International Finance. (4th press). New York: Palgrave Macmillan. Quantitative Easing (QE) – Injecting Money Into the Economy. (n.d.) Retrieved from

http://www.bankofengland.co.uk/education/Documents/resources/postcards/qecomp.p df

Swanson, E. T., Reichlin, L., & Wright, J. H. (2011). Let's Twist Again: A High-Frequency Event-Study Analysis of Operation Twist and Its Implications for QE2 [with

Comments and Discussion]. Brookings Papers on Economic Activity, 151-207. Tillmann, P. (2015). Unconventional Monetary Policy and the Spillovers to Emerging

Markets. Retrieved from

https://www.cb.cityu.edu.hk/ef/doc/2015%20Conference%20on%20the%20New%20 Normal/Peter%20Tillmann(1).pdf

João Barata R.B. Barroso, Luiz A. Pereira da Silva, Adriana Soares Sales, Quantitative easing and related capital flows into brazil: measuring its effects and transmission channels through a rigorous counterfactual evaluation, Journal of International Money and

Finance (2015), http://dx.doi.org/doi:10.1016/j.jimonfin.2015.06.013.

Serial Correlation, the Durbin-Watson Statistic, and the Cochrane-Orcutt Procedure. (2006, February 26). Retrieved from

http://www.albany.edu/faculty/kretheme/PAD705/SupportMat/SerialCorrelation.pdf Thomas, L. (2014, January 28). ‘Fragile Five’ Is the Latest Club of Emerging Nations in

Turmoil. The New York Times. Retrieved from

http://www.nytimes.com/2014/01/29/business/international/fragile-five-is-the-latest-club-of-emerging-nations-in-turmoil.html?_r=0

(24)

Appendix

Brazil

Regressor (1) (2) (3) (4) (5)

QE 5.13e-07* 5.98e-07* 6.80e-07* 3.30e-07 3.31e-07

IR 0.1401585* .1557078* .170533* .1706814* EXR 0.992649* 2.275562 -0.2707511 UnemR -.7385347* -.7352729* RAW 0.0002062 Constant 4.407582 2.755234 3.900619 8.956939 8.895286 Root MSE 0.3087 0.85356 0.83977 0.83551 0.84172 Adj. R2 0.0101 0.3481 0.369 0.4537 0.4456

Fig. 5 OLS regression output Brazil South Africa

Regressor (1) (2) (3) (4) (5)

QE -7.43e-07* 8.98e-07* -4.98e-09 -3.27e-08 -1.33e-06*

IR .9543971* .6555983* .747807* .2903912* EXR .5428747* .5194808* 1.045153* UnemR .2641436 0.082254 RAW .0464355* Constant 7.77747 -2.406531 -2.970118 -9.743825 -7.854987 Root MSE 1.6529 1.1014 1.0079 1.0072 0.73745 Adj. R2 0.2065 0.6477 0.705 0.7054 0.8421

(25)

India

Regressor (1) (2) (3) (4) (5)

QE -4.54e-08 6.61e-08 -2.95e-07 -.0000265* -.0000258*

IR -.1130093 0.4584753 1.292879* 1.358652* EXR 0.13455 0.3012122 0.0845365 UnemR 91.28197 113.8722 RAW -0.069094 Constant 9.200735 9.77247 6.24876 -290.6222 -358.7531 Root MSE 2.4921 2.5033 2.4839 1.4915 1.5067 Adj. R2 -0.0101 -0.0191 -0.0034 0.6634 0.6565

Fig. 7 OLS regression output India Indonesia

Regressor (1) (2) (3) (4) (5)

QE -5.51e-07* 1.62e-07 8.81e-08 -8.00e-07* -9.81e-07*

IR 1.473542* 1.413824* 3.021671* 2.469933* EXR 0.0001003 -0.0005319* 0.0001395 UnemR -2.597975* -1.855728* RAW 0.0507225* Constant 7.217584 -4.827401 -5.234113 11.14203 -1.11417 Root MSE 2.1336 1.5934 1.5999 1.2101 0.91023 Adj. R2 0.0733 0.4831 0.4789 0.7225 0.843

(26)

Brazil

Regressor (1) (2) (3) (4) (5) (5), corc

QE 5.68e-07* 5.05e-07* 5.03e-07* 5.35e-07* 5.36e-07* -7.98e-07*

IR 0.1752291* .174979* .2860804* 0.2761651* .151783* EXR 0.0142132 0.3707124 0.2930516 0.0067513 UnemR -0.18729 -0.2178211 0.0590524 RAW -0.0018038 0.0066355 Lag1 -0.1552646 -0.10613 -0.1085006 -0.277253 -0.2921681 -.3834355* Lag2 0.1178229 -0.0522 -0.0502521 -0.296559 -0.258141 -0.1244397 Lag3 -0.4071003 -0.05927 -0.0596981 0.424132 0.3785464 -0.0896755 First.d 1.129445* 0.6776592* .6706129* 0.0997003 0.0887701 0.2081053 Tapering.d -0.5712648* 0.4562533-0.4590823* -0.877787* -.8403551* 0.2701209 Constant 6.837891 3.975121* 3.957835 3.276515 4.016968 9.788702 Root MSE 0.62603 0.58871 0.59267 0.52207 0.52714 0.20605 Adj. R2 0.4107 0.4858 0.4858 0.6372 0.6377 0.3247

Fig. 9 Dynamic OLS regression output Brazil South Africa

Regressor (1) (2) (3) (4) (5) (5), corc

QE -8.32e-07* 6.48e-07* -3.10e-08 1.89e-07 -4.24e-07 -5.02e-07

IR .9380928* .7284426* .2959895* .2230118* 0.2231184 EXR 0.3773017* 0.3882509* .6476175* 0.2014355 UnemR -1.547849* -1.099971* -.4686312* RAW .0220852* .0171858* Lag1 1.026684* 0.6264567* 0.324953 0.3613006 0.3171729 -0.17966 Lag2 -0.2078633 -0.1260684 -0.038121 -0.2729093 -0.2057496 -0.1098083 Lag3 -.9251119* -0.5676022* -0.4334715 -0.2461419 -0.2023126 0.1036212 First.d 1.96839* -.1347974 -.6212674 -0.7597249* -.3412267 0.2759427 Tapering.d -.0126441 -.4160061* -.5218479* -1.044862* -1.066015* -0.248428 Constant 8.783238 -1.082449 -0.8747067 39.28057 25.61535 15.31028 Root MSE 1.5335 0.88725 0.85974 0.72422 0.68052 0.36795 Adj. R2 0.4362 0.8138 0.8275 0.8792 0.8948 0.1395

(27)

India

Regressor (1) (2) (3) (4) (5) (6) (5), corc

QE -6.51e-07* -7.51e-07 -6.10e-07 -.0000168* -.0000167* -2.03e-07 .0000118

IR .0727123 .121642 1.266155* 1.198695* 0.5903428 -.120636 EXR -0.0324674 0.2172495 0.408799 -0.1735523 0.0323281 UnemR 15.3938 -10.69126 - -108.7877 RAW 0.0616482 -0.0435195 0.0220521 Lag1 -.0088191 .0003578 -.0063505 -.6621233* -0.7588014* -0.0314241 -.5881109* Lag2 -0.1578584 -0.1555293 -0.1493378 0.0010481 .0503856 -0.1809633 -0.0593149 Lag3 0.1847414 0.1965665 0.1681973 -.1767612 -.1602558 0.2090355 -0.2636038 First.d -0.4340132 -0.4376393 -0.2838239 - - -0.7509335 -Tapering.d 0.4329046 0.4496453 0.541414 - - 1.201112 -Constant 10.87848 10.35463 11.61532 -20.72725 60.67163 18.53444 390.199 Root MSE 2.3422 2.3571 2.3715 1.0625 1.0711 2.2857 0.72939 Adj. R2 0.0902 0.0907 0.0918 0.8812 0.8868 0.1677 0.4418

Fig. 11 Dynamic OLS regression output India Indonesia

Regressor (1) (2) (3) (4) (5) (5), corc

QE -8.76e-07* -2.74e-07* -2.99e-07* -3.36e-07 -6.99e-07* -5.45e-07

IR 2.058649* 2.034445* 2.723697* 1.928774* 1.822327* EXR 0.0000339 -0.0003026 0.0004731* .0004155* UnemR -1.078452* -0.4496785 .0138669 RAW 0.0400422* .0348999* Lag1 0.19841 -0.3635923 -0.371094 -0.0210878 -0.1700387 -.2945395* Lag2 -0.25231 -0.1592421 -0.152435 -0.2794854 -0.1511072 -.1389044 Lag3 -0.41572 0.0026596 -0.001568 -0.0667533 -0.0950976 -.0260555 First.d 4.512204* .1107357 .0662251 0.216933 .5030674* -.0204905 Tapering.d 2.329107* .0207793 -.0101511 -.4218712 -1.011969* -.5076476 Constant 10.7267 -4.521305 -4.603598 1.111283 -8.110164 -9.823057 Root MSE 2.0219 0.88138 0.887 0.9164 0.52477 .42004 Adj. R2 0.3333 0.875 0.8751 0.77387 0.9622 0.8387