Login Get Started
HelpDesk
 Embed  Download
  of  

or
Sign in to continue reading...

The Importance of External Shocks and Global Monetary Conditions for A Small-Open Economy

Gulnihal Tuzun
By Gulnihal Tuzun
3 years ago
The Importance of External Shocks and Global Monetary Conditions for A Small-Open Economy

Banking

International Trade

Organisation Tags (2)


Create FREE account or Login to add your comment
Comments (0)


Transcription

  1. The Importance of External Shocks and Global Monetary Conditions for A Small-Open Economy G ülnihal Tüzün April 2021 Working Paper No: 21/09
  2. © Central Bank of the Republic of Turkey 2021 Address: Central Bank of the Republic of Turkey Head Office Structural Economic Research Department Hacı Bayram Mah. İstiklal Caddesi No: 10 Ulus, 06050 Ankara, Turkey Phone: +90 312 507 80 04 Facsimile: +90 312 507 78 96 The views expressed in this working paper are those of the author(s) and do not necessarily represent the official views of the Central Bank of the Republic of Turkey.
  3. The Importance of External Shocks and Global Monetary Conditions for A Small-Open Economy G ülnihal Tüzün∗ Abstract The purpose of this study is to assess how do the domestic and foreign shocks affect the fundamental macroeconomic variables of a small-open economy, and in particular Turkey. The domestic supply, demand and monetary policy shocks as well as their global counterparts are identified by employing a Bayesian structural VAR model with sign and zero restrictions. After a US monetary tightening shock, the results demonstrate an appreciation of US Dollar against Turkish lira, a rise in the consumer price level in the Turkish economy, a contractionary monetary policy shock accompanied by a fall in the real output level. This reaction is a strong evidence of the existence of a global interest rate contagion present in the international macroeconomics literature. JEL classification: C11, C32, E52, F41 Keywords: Bayesian VAR; sign and zero restrictions; shock identification; monetary policy. ∗ Central Bank of the Republic of Turkey, Research and Monetary Policy Department. Email: gulnihal.tuzun@tcmb.gov.tr Author thanks to the anonymous referee and the participants of the CBRT Brown Bag seminar for the constructive recommendations and discussions. Author is also grateful to Hande Küçük, Tayyar Büyükbaşaran, Gökçe Karasoy-Can and Selçuk Gül for insightful comments and suggestions which help improve the final version of the paper. All errors are mine. The views expressed in this paper are those of author and do not necessarily reflect the official views of the Central Bank of Turkey or its staff.
  4. Non-Technical Summary With the current level of economic globalisation , it is of utmost importance both for economists and policy makers to identify the existing international macroeconomic connections between the advanced and emerging market economies. Therefore, with the ever increasing macro-financial linkages around the globe, studies pertaining global spillovers have augmented their space in international and open-economy macroeconomics literature. This paper attempts to answer two main questions related to the international macroeconomic shock transmission mechanisms in the extent of the Turkish economy: To what extent are the fluctuations in an emerging market caused by the US monetary policy? What is the role of the global supply and demand shocks on key macroeconomic indicators and how those shocks are transmitted to emerging market economies? Accordingly, the paper aims at identifying the following external shocks: foreign monetary policy, oil price, global aggregate demand and risk premium. After a US monetary tightening shock, the results demonstrate an appreciation of the US dollar. This in turn causes a rise in the price level. Monetary policy in Turkey demonstrates a tightening stance. In the end real output level falls through the monetary policy transmission mechanism. As the exchange rate pass through explains the currency depreciation and its repercussion on pricing dynamics, the contraction in the economy is attributed to the balance sheet and the interest rate channels. Moreover, after a US monetary tightening, a tightening reaction by the Central Bank of Turkey (CBRT) is observed as an anticipated monetary policy action as a small, follower economy. This reaction is in line with the existence of a global interest rate contagion present in the international macroeconomics literature. A conditional scenario analysis is conducted while the expected federal funds rate policy path constitutes the conditioned shock as the foreign monetary shock. After a foreign monetary easing scenario for the second half of 2019, a significant easing of the CBRT policy rate, a fall in the consumer price level, a relatively stable exchange rate and an upward trend in the real domestic output for 2020 are observed. Apart from the analyses of the impulse response functions, the forecast error variance decomposition analysis reveals that while the global shocks explain a substantial part of the variation in domestic policy rate and the bilateral exchange rates, domestic shocks do still 1
  5. contribute to these variables . Besides, the external and domestic shocks have approximately the same information for explaining the forecast error variance of those variables at different horizons during the estimation period. 2
  6. 1 . Introduction The current level of the economic globalisation has reached a stage that includes a considerable degree of economic and financial interconnection among advanced and emerging economies. As the empirical analyses provided a substantial evidence of the existence of a global financial cycle whose main determinants are the monetary policies conducted by the core economies (Rey, 2015), it is worth examining advanced economies’ macroeconomic policies and their repercussions for small open economies. There are central questions regarding the fluctuations in the emerging market economies that have been open to debate for many decades. These can be specified but are not limited to the following: To what extent are the macroeconomic fluctuations in emerging markets originated from abroad? In particular, considering the fact that the developments in the United States (US) economy have sizeable effects, to what extent are the fluctuations in emerging markets caused by the US monetary policy? What is the role of the global supply and demand shocks on macroeconomic indicators, connection of these shocks with the US economy, and how they are transmitted to the rest of the world? Therefore, evaluating the economies of the emerging markets independent from the rest of the world and engaging in macroeconomic policy making in isolation are inappropriate. Thus, the discussion in this paper demonstrates the importance of the consideration of emerging economies together with the rest of the world while making a sound macroeconomic policy design. In this framework, Turkey appears to be an appropriate candidate country to examine along the lines of these existing issues due to several reasons. Turkey has a representative, exemplary emerging market economy to delve into due to its exposure to the external shocks. As an oil-importing open economy, external developments such as oil price or foreign monetary policy shocks constitute a major input for macroeconomic policies due to external shocks’ direct impact on the economy’s dynamics. At the same time, Turkey is as well sufficiently small that allows us to mute the impact of domestic shocks on foreign shocks. Thus, it is a natural laboratory to apply the block exogeneity restrictions that lie at the core of the models for small-open economies. In that sense, the model set up can be adapted to other emerging market economies while bearing the country-specific fine tuning in mind. In this paper, foreign monetary, global supply, global demand and the domestic supply and demand shocks are identified. With such a system in which domestic and foreign shocks are identified simultaneously, the aim of this paper is to shed light on the international shock transmission channels on macro variables. To this end, an empirical approach is preferred and a sign and zero restricted Bayesian Structural VAR model is estimated to answer these empirical questions. 3
  7. The results are as follows . After a US monetary tightening shock, the results demonstrate a significant appreciation of the US dollar against Turkish lira. This in turn causes a rise in the price level. Monetary authority responds with a tightening policy. In the end real output level falls through balance sheet channel and interest rate channel as these channels outweigh the role of the expenditure switching channel. These findings are consistent with what the previous studies on international macroeconomics literature documents for the impact of the US monetary policy on emerging market economies (see Dedola, Rivolta, and Stracca (2017); Maćkowiak (2007); Canova (2005) among others). In particular, the price level increases due to the impact of the exchange rate pass through. In response to that, as a small, follower economy, an emerging market’s domestic monetary policy stance becomes contractionary.1 This reaction may be interpreted as an obvious evidence of the existence of a global interest rate contagion discussed extensively in the international macroeconomics literature. This paper contributes to the literature on international monetary policy spillover which has evolved around panel models where country-specific characteristics are taken into account or two-country time-series models. Our study rests on the second strand of the literature, on a two-country model where the emerging market in concern is Turkey that demonstrates periphery economy characteristics. On the other hand, the rest of the world is regarded as the core economy that exhibits the features of an advanced economy. A substantial amount of literature exists on the international shock transmission mechanism of the US monetary policy. However, reviewing the impact of global supply and global demand disturbances along with the US monetary shock on the Turkish economy has remained a gap in the literature. This paper fills this gap by incorporating both domestic and foreign shocks in a single Bayesian VAR model with sign and zero restrictions. There are studies examining the global and domestic shocks in recent international macroeconomics literature. These studies primarily rely on vector autoregressive (VAR) models with Bayesian estimation to overcome the (curse of) dimensionality issue (such as Jovičić and Kunovac (2017) for Croatia, Szafranek, Hałka, et al. (2017) for Poland, Conti (2017) for the US, Conti, Neri, and Nobili (2017) and Bobeica and Jarocinski (2019) for the Euro area). While some studies in the literature (such as Dedola et al. (2017); Georgiadis (2016)) extract the US monetary policy shocks à la Gertler and Karadi (2015) and then regress the macroeconomic variables on the foreign monetary policy surprise series, others rely on two-country structural VAR models. What is central in those set of studies is that they use minimum set of restrictions to identify the domestic and foreign shocks. The aforementioned papers, as adopted in our study, predicate on either empirically or theoretically plausible restrictions 1 Hofmann and Takáts (2015) argue that more than one-to-one relationship exists between the Federal Reserve (FED) Funds Rate and policy interest rates of the emerging market economies. 4
  8. in their identification schemes . However, it must be carefully noted that the dynamics in these economies differ substantially from those of Turkey due to Turkey’s being an emerging market economy. Thus, apart from the country-specific factors, expert opinion-based sign restrictions are employed as well so as to identify the related shocks in this paper. The line of reasoning behind them are explained in detail in Section 3. Emerging markets have also been investigated by previous studies (see Canova (2005); Maćkowiak (2007); Hajek and Horvath (2018); Dedola et al. (2017); Georgiadis (2016), among others) in terms of their exposure to the external shocks. Results reveal that the response of output in periphery countries after a US monetary policy shock is mixed. However, domestic currency depreciation and a subsequent rise in consumer prices are common outcomes after a contractionary foreign monetary policy. Buyukbasaran, Can, and Kucuk (2019) identify domestic shocks for Turkey in a Bayesian structural VAR setting from a credit supply perspective. However, our study differs from rest of the studies by incorporating global shocks into the model. Apart from the aforementioned papers on advanced economies, this study is the first in the literature that covers both the domestic and foreign shocks simultaneously for a small-open economy, to the best of our knowledge. Furthermore, more precise and robust results have been achieved in terms of the shock transmission channels. The remainder of the paper is structured as follows: Section 2 presents the empirical model and the algorithm, Section 3 introduces the identification strategy used in the models of the paper and reports the results, Section 4 provides some alternatives for robustness checks and Section 5 summarizes the findings and concludes. 2. Empirical Methodology 2.1. Model The structural vector autoregressive (VAR) models have been the workhorse of empirical macroeconomics and finance over the last decade. SVARs incorporate additional identifying assumptions motivated based on economic theory on model responses in contrast to the standard, reduced-form VAR representations. Pinning down from a structural form to reduced form requires certain assumptions about the matrix which establishes the relation between the reduced form residuals and the structural disturbances (innovations). This is the identification issue central in the questions in this research. Consider the reduced form VAR(p) model: yt = A1 yt−1 + A2 yt−2 + ... + Ap yt−p + εt 5 (1)
  9. with εt ∼ N (0, Σ). Consider the structural form of the above VAR model. B0 yt = B1 yt−1 + B2 yt−2 + ... + Bp yt−p + ηt (2) with ηt ∼ N (0,Γ) is the vector of structural disturbances with an orthogonal Γ matrix. Define, B = B0−1 (3) Pre-multiplying both sides of the Eqn.2 yields us the following: εt = Bηt (4) Here, the objective is finding B, which is the structural impact matrix. However, there are infinitely many ways to decompose the diagonal Γ matrix into B and B , most common way of which is either recursive (Cholesky) identification scheme or imposing the sign restrictions approach. The reduced-form VAR does not contain any information regarding the matrix B, and because of the unknown B, estimates of the Eqn.1 can not be used to identify the structural error terms. By using Eqn.4 one can formulate the IRFs of the structural VAR model with a diagonal structural impact matrix which can be interpreted with an economic line of reasoning. In this setting, the main motivation is to assess how the main macroeconomic and financial aggregates of the Turkish economy, i.e domestic output, price level, exchange rate as well as the domestic policy interest rate are affected from the global monetary, supply and demand conditions. Quarterly data from 2003:II to 2018:IV is used. The data set is composed of two blocks of variables that can be partitioned into the foreign and domestic blocks. As a representative of the global monetary conditions, the effective FED funds rate is preferred throughout this study. As the changes in US monetary policy has a sizeable crossborder impact on global financial markets, the choice for the FED funds rate is deemed appropriate. After the Global Financial Crisis, when the effective FED funds rate hit the Zero Lower Bound (ZLB), to stimulate the economy and to be able to recover from the disinflationary period, FED involved in unconventional policy measures through Large Scale Asset Purchase programs forward guidance as main tools of the Quantitative Easing (QE) period. Therefore, for the period between December 2008 to December 2015, when the effective FED funds rate fluctuates between the level zero and 25 basis points (bps), it is inappropriate to use the effective FED funds rate as it does not represent the actual monetary 6
  10. Figure 1 . Comparison of Effective FED funds rate and Shadow Rates Source: FRED Economic Data, CBRT, Bloomberg policy stance which was looser than the level of the effective FED funds rate suggests. Thus, in order to capture the unconventional policies and to quantify the stance of the monetary policy, it is vital to use the corresponding shadow rates to the QE period (see the Figure 1). These shadow rates have been argued to successfully approximate the monetary policy stance and to account for the ZLB period. The shadow rates calculated by the Wu and Xia (2016) and the Krippner (2013) are employed in this paper and they offer an approximation to a term structure model that is tractable for an analysis of the US economy operating near the zero lower bound. To be used in the baseline model, two other global variables were added into the system: the oil prices and the world industrial production volume (excluding construction) as a proxy for global demand. For the domestic block, the main macroeconomic aggregates of Turkey are used. The real GDP, the price indicator as to the consumer price index, USD/TRY nominal exchange rate and the policy rate for CBRT2 are used in the domestic block. For the variables except for the financial ones, "light transformation" is adopted to achieve stationary series and to achieve stability condition in VAR models. All variables are seasonally adjusted (except the exchange rate and the interest rates) and log-differenced to achieve stationarity, except the FED funds rate and policy rate which entered into the model as levels. The full description of the variables, including the list of the series, the source of the data and the 2 Prior to May 2010, CBRT overnight borrowing rate and after 2010 Q2 the Istanbul Stock Exchange (BIST) overnight borrowing rate is used. 7
  11. applied transformations where used necessary are provided in the Appendix D . 2.2. Block Exogeneity Feature The identified shocks used in the Initial Model are as follows: domestic monetary/aggregate supply/aggregate demand, risk premium and foreign monetary shock. In the Baseline Model, additional global supply and demand shocks are articulated into the initial model so that a complete macro-model framework is achieved by building on the first scheme. The assumption of domestic variables have an impact on the foreign ones after the initial period is unrealistic as an emerging market economy can not have an impact on an advanced economy. Therefore, the domestic variables need to be exogenous as a block to the foreign ones. p s=0 A11 (s) A12 (s) A21 (s) A22 (s) y1 (t-s) c11 ε1 (t) + = y2 (t-s) c21 ε2 (t) (5) In the above equation, y1 (t) represents the vector of macroeconomic variables of the global economy, i.e. the external world, y2 (t) represents the variables of the small economy, and vectors c11 , c21 are constants. Finally, ε1 (t) and ε2 (t) denote structural shocks of the rest of the world and the domestic country. For each A12 (s)=0, the variables of the emerging market economy are determined to be exogenous to the variables of the rest of the world such that the neither current nor the past (the lagged) developments in small-open economies can not affect the rest of the world. This assumption is called as the block exogeneity 3 feature as the foreign variables are exogenous to those of the small open economy. Apart from the block exogeneity restrictions, in order to avoid puzzles4 due to the drawbacks of recursively ordered semi-structural schemes, sign restrictions constitute an integral part of the identification. To avoid improper identification of the system of VAR, building models by using Structural VARs with sign and zero restrictions is preferred in this study. The line of reasoning and the identification schemes using the sign restrictions are elaborated in detail at Section 3 below. 3 This feature has been introduced firstly by Cushman and Zha (1997) and used later as well by the others: Canova (2005), Maćkowiak (2007) and Dungey and Pagan (2009) who investigate the impact of US shocks on Latin American countries, on East Asia-Latin American emerging markets and on Australian economy, respectively. 4 The most common puzzles in the open macroeconomics literature are the price, liquidity and exchange rate puzzles such that all of which results from encountering contradictory responses of variables after a monetary tightening that are at odds with theoretical, anticipated responses. 8
  12. 2 .3. Algorithm In this paper, estimations are carried out with the macro-modelling BEAR Toolbox developed by Dieppe, Van Roye, and Legrand (2016) with the following algorithm. Consider the Eqn.1 i.e., the reduced form VAR model and its residual covariance matrix Σ. The Ψi denote the impulse response functions obtained from the reduced VAR and let h(Σ) be a preliminary structural matrix, where h(Σ)×h(Σ) = Σ. From this matrix, obtain a set of structural impulse response functions, Ψi for i =0,1,2..; Ψi = Ψi h(Σ) (6) To draw from the correct posterior distribution and to implement an orthogonalisation step, draw a random matrix Q from a uniform distribution and define, B = h(Σ)Q (7) The aim is to draw such an orthogonal Q to preserve the SVAR property; BΓB’ = BIB’ = BB’ = h(Σ)QQ’h(Σ ) = h(Σ)Ih(Σ ) = h(Σ)h(Σ ) = Σ (8) Σ = BΓB’ (9) Now we need to obtain an orthogonal matrix Q from the uniform distribution. To do so, first draw a n × n random matrix X from an independent standard normal distribution. Then use QR decomposition of X, such that X = QR, with Q an orthogonal matrix and R an upper triangular matrix. Considering the structural impulse responses as: ˜ = ΨB = Ψh(Σ)Q = Ψi Q Ψ (10) and the stacked structural matrix, when the restrictions are implemented for periods p1 , p2 ,..., pn , then f(B, B1 ,...,Bp ) can be written as:    ˜ p1 Ψ Ψp1 ˜     Ψp2   Ψp2       =  .  Q = f (B,B1 ,...,Bp ) × Q f(B, B1 ,...,Bp ) =  .          .   .  ˜ pn Ψ Ψpn  9 (11)
  13. where  Ψp1    Ψp2     f (B,B1 ,...,Bp ) =  .      .  Ψpn  (12) Here, if the restrictions hold, then the Sj × fj (B, B1 , ..., Bp ) > 0 (13) is satisfied for all the shocks j = 1,2,..,n ; where fj (B, B1 ,...,Bp ) represents the j th column of the matrix f (B, B1 ,...,Bp ) and where Sj is the restriction matrix with a number of columns equal to the number of rows of f (B, B1 ,...,Bp ) and a number of rows equal to the number of sign restrictions on shock j. Then keep the matrix Q and go for the next iteration. If the condition (13) does not hold, repeat the process from drawing the reduced form coefficients, and Σ and continue the whole algorithm until a valid Q matrix is obtained. The algorithm for the above procedure is provided at the Appendix, Section A. Throughout the algorithm of Arias, Rubio-Ramirez, and Waggoner (2014), Bayesian inference is used for the determination of the posterior distribution of the two blocks in the VAR system, i.e the coefficients of the variables, β’s, and the variance covariance matrix, the Σ, as it is an appropriate methodology for dealing with small-sample properties, while restricting these blocks with a-priori restrictions. 3. Identification & Results 3.1. Initial Model The initial scheme consists of five shocks originated from domestic and foreign conditions. Apart from the first three domestic shocks, the fourth shock is for introducing a global monetary shock represented by the effective FED funds rate (and shadow rates where necessary). The last one represents a (positive) risk premium shock. In the models, instead of imposing the signs for longer horizons, as the seminal paper of Uhlig (2005) opts as an identification strategy, the (sign and zero) restrictions are put only for the first period. This is to observe how the variables respond and let the model dynamics determine the response’s behavior after the initial period. 10
  14. Table 1 : Shocks and Restrictions for the Initial Model Shock/ Variables FED Funds Rate $/TL Exchange Rate Domestic Consumer Prices Domestic Real GDP CBRT Policy Rate Domestic Monetary Domestic AS Domestic AD 0 + 0 0 + • 0 • + + + Foreign Monetary Risk Premium + • • • + • + • Notes: • = no restriction, + = positive sign, - = negative sign. All restrictions are imposed only on impact, the zeros denote the block exogeneity when the associated shocks are domestic ones due to the block exogeneity assumption. The exchange rate is defined as a "-" indicates an appreciation. A positive domestic aggregate supply shock (hereafter the domestic AS) is the shock in the system that moves the domestic real GDP and consumer prices in opposite directions. In that context, the source of the domestic supply shock can be considered as a total factor productivity shock. Because there exists no systematic and optimal monetary policy reaction to the supply shocks existing in the economy, the response of the policy rate to the AS shock is left agnostic. To disentangle the domestic AS shock from a risk premium shock in the model, the response of the exchange rate variable is restricted for the first quarter in response to a domestic AS shock. These restrictions are imposed following the rationale related to the domestic consumption and investment behaviours in New-Keynesian models documented by Smets and Wouters (2003). A positive domestic demand shock moves both the real GDP and consumer prices in the same direction. Also, the response of the policy rate to a domestic aggregate demand (hereafter the domestic AD) shock is positive5 and it operates through the CBRT’s monetary policy reaction function. A positive domestic monetary shock is defined as a contractionary monetary policy shock such that it decreases the consumer prices and real GDP while appreciating the exchange rate. It must be noted that for all the domestic shocks, they have no impact on FED funds rate. It stems from the fact that Turkey is a small emerging market economy and shocks originated from the domestic economy is anticipated to have no impact on the monetary policy of the FED. The foreign monetary shock is a contractionary FED monetary policy shock which induces central bank of the domestic economy to increase the policy rate. The line of reasoning behind this assumption rests on the empirical findings from the international monetary policy spillover literature. Hofmann and Takáts (2015) find theoretically and statistically significant 5 As a robustness check, we left the policy rate response to the domestic AD shock unrestricted and still, there is a positive reaction of the domestic policy rate to the demand shock. 11
  15. monetary policy spillovers that originate from the US to emerging market economies . They argued the presence of an international interest rate co-movement in recent years despite the considerable variation in business cycles among countries. Their paper reported that a 100 bps change in the federal funds rate is associated with between a 26 bps to 46 bps shift in the policy rate of the emerging economies that is significant both economically and statistically. Likewise, in a panel VAR framework of 43 emerging and advanced economies, Caceres, Carriere-Swallow, Demir, and Gruss (2016) find that a 100 bps increase in US short term interest rate leads to a response of approximately 20 basis points rise in domestic shortterm interest rates in rest of the world. Figure 10 at Appendix D demonstrates that central bank policy rates of emerging economies follow closely that of the US. The above-mentioned empirical studies point out a consistent co-movement of interest rates across countries. This is primarily due to the central banks’ response to macroeconomic developments such as inflationary pressures originated from either supply or demand side conditions, exchange rate fluctuations to a certain extent and labor market conditions depending on their mandates. In that respect, the Figure 11 at Appendix D depicts that although this historical relationship has weakened for some period in time, there is a positive co-movement between the FED funds rate and that of the CBRT. As a small economy, the policy rate of the CBRT is anticipated to follow the USA policy rate. Thus, after a positive US monetary policy shock (a contractionary FED funds rate, ffr ), domestic monetary policy responds positively. Rest of the variables following this shock are left agnostic. The last shock is a risk premium disturbance. It steps in the model as a positive risk premium factor for emerging markets such that it lowers the consumer prices, increases the output level while appreciating the exchange rate. The FED funds rate variable is left unrestricted to a risk premium shock as it is considered as a positive risk premium shock to an emerging economy while depreciating the US dollar. In that sense, the FED may (or may not) respond to this negative exchange rate shock. With these sets of signs and restrictions described above, all of the shocks in the structural system are identified and disentangled from each other. The priors and the related hyperparameters that are used to estimate the structural model are provided at Appendix C: The resulting impulse-response graphs are as follows for the initial model. 12
  16. Figure 2 . Initial Model’s Responses Note: The dark blue line represents the median of the posterior distribution. The shaded light blue area is the 0.68 probability interval of the posterior distribution. STN denotes the CBRT policy rate variable. The domestic monetary policy shock satisfies all the imposed restrictions as can be seen from the Figure 2. Following a contractionary domestic monetary policy, the price level and output responds in the same direction with narrow credibility bands, while Turkish lira appreciates. As a small-open economy, a monetary policy shock originated from Turkey has no impact on the monetary policy stance of the FED. The frequently observed puzzles in the literature do not appear in the structural model results with identified shocks. One thing to note that, the shaded areas should not be interpreted as confidence bands (or intervals) as conventionally regarded in VAR models estimated with ordinary least squares. The bands in the SVAR models are the Bayesian credibility intervals showing the 16th and 84th percentiles of the whole accepted draws from the posterior distribution. Regarding the domestic AS shock, it causes price level to decrease and output to increase. As can be seen from the figures, the exchange rate does not respond significantly throughout the response horizon, except for the first period upon which the restriction is imposed. It should be noted that although monetary policy of Turkey manifests an accommodative stance after three periods, it is not significant despite the larger portion of the confidence bands as well as the median response falling below the zero line. A domestic AS 13
  17. shock can be considered as a positive total factor productivity shock or a permanent negative risk premium shock where the monetary policy stance is anticipated to stay neutral . Therefore, empirical results suggest that there is not optimal monetary policy response to (aggregate) supply shocks. In case of a domestic AD shock, the price level and output reacts in the same direction and as a monetary policy reaction to that the domestic policy interest rate increases. Following that, the exchange rate displays a limited appreciation, although of an insignificant type but still skewed to the lower portion of the zero line up to two quarters. It is widely found as an empirical evidence in the literature that, after a US monetary policy contraction, the exchange rates of the emerging markets depreciate (Gupta, Masetti, and Rosenblatt, 2017). After the US monetary contraction, it is anticipated that the capital outflows from the rest of the world increases and the demand for emerging market currencies declines. Because the actors in the economy seek assets giving higher yields, they allocate their portfolios towards higher return assets. In that framework, US dollar appreciated approximately 3.8 per cent on impact after a 155 bps FED funds rate hike. Model results suggest that following the global monetary shock, the domestic policy rate increases, around 360 bps, more than the FED increases its main policy rate (this amounts to 60bps policy rate rise following a 25 bps FED funds rate rise with an approximately 0.6 per cent US dollar appreciation). In that sense, a more than a one-to-one response to global monetary policy shock is observed in the Turkish economy. As a consequence, there is 0.55 percentage points of contraction in real GDP. After a FED policy rate hike, significant price level rise is observed (around 0.22 percentage points) as well as a fall in real GDP. These results are in line with other studies focusing on the international spillover of US monetary policy such as Maćkowiak (2007), Caceres et al. (2016), and Demir (2019) where they find a contraction in real GDP and a rise in interest rates after a US monetary contraction. In case of a positive risk premium shock, the price level falls and output demonstrates a hump-shaped behaviour and rises to first two quarters then starts to fall as the FED funds rate is increased too. In case of a risk premium fall for emerging markets, a contractionary monetary policy stance is observed in the US, which makes sense considering the positive risk premium (RP) shock as an unfavourable circumstance for the value of the US dollar. The essential thing regarding the identification of the structural shocks of the initial model is that although a full set of domestic shocks are present in the model, identifying solely the US monetary policy shock and the risk premium shock as external shocks does not seem to be sufficient enough to capture the structural disturbances relevant for the small-open economy. The line of reasoning behind this reservation is as follows: 1. Whether the FED responds to its domestic demand conditions, or 14
  18. 2 . It responds to the global supply/demand conditions driven either by the oil market developments or global real activity is unclear. 3. Global aggregate supply and demand shocks are required to be incorporated into the system in order to complete and close the system in a macroeconomic modelling perspective. Due to the aforementioned reasons, adding two more shocks into the system to fully identify the external shocks is presented in the Baseline model. 3.2. Baseline Model Adding additional oil price and global aggregate demand shocks will complete the system and it will enable to identify both domestic monetary, supply and demand shocks with their global counterparts related to Turkey as an emerging economy. Global AD shock is thought as a standard demand shock in accordance with a positive change in global real activity. The oil price shock will operate in the system as a negative oil supply disturbance which contracts the overall economic activity in the world. It is considered as a rise in oil prices due to a decrease in oil supply. In that sense, it could be thought as a global negative aggregate supply shock. With these additional disturbances, the main macroeconomic aggregates in Turkey as an emerging economy will be fully identified with the associated macro-shocks by this approach. Firstly, global demand shock is introduced in the model in order to partial out a possible reaction of the FED to global demand conditions. However, it appeared as an insufficient improvement, as a global supply shock is needed to be disentangled from a candidate global AD shock from an oil price shock with regards to the identification. Both global supply and demand shocks are included simultaneously to assess how the emerging market economy in concern responds to each shock. The identification scheme regarding the full model is as follows: 15
  19. Table 2 : Baseline Model Identification Shock / variable Domestic Monetary Domestic Aggregate Supply Domestic Aggregate Demand Risk Global Premium Aggregate Demand Oil Price Foreign Monetary Global Demand Oil Price FED Funds Rate Domestic Consumer Prices Domestic Real GDP CBRT Policy Rate $/TL Exchange Rate 0 0 0 + - 0 0 0 + • 0 0 0 0 + + + • 0 0 • + • - + 0 + • • • • • + • • + • + + • + + • • Notes: • = no restriction, + = positive sign, - = negative sign. All restrictions are imposed only on impact, the zero’s denote the block exogeneity when the associated shocks are domestic ones due to the block exogeneity assumption. The exchange rate is defined as so that a - means an appreciation. The domestic supply, demand and monetary shocks, as well as the US monetary policy shocks, are preserved in the model as explained in previous subsection. The same priors and the hyperparameters are preferred as given in the Appendix. The strategy for identification in this baseline model mostly follows the logic of Conti et al. (2017) and Bobeica and Jarocinski (2019). Identification of the global shocks is as follows6 : An oil price shock is assumed to channel into the system as a negative global aggregate supply shock. An exogenous increase in the real price of oil causes the overall price level to rise as it is used both as an input for production and consumption. Because the oil price rise in question is thought to originate from a negative global supply shock, it decreases the global output. With the negative sign for global demand, oil price shock is disentangled from the global demand. The response of output level of Turkey to a negative oil supply shock is left unrestricted because of the following: Output level of Turkey may fall as the global economic activity contracts after a positive oil price shock. On the other hand, because of higher oil prices, the export partners’ of Turkey that are oil exporters (for instance the OPEC members) can experience a positive wealth effect, thereby boosting their imports. This could generate increased exports for Turkey and may imply a rise in the output level for Turkey, too. Thus, the response of the real GDP of Turkey to the positive oil price shock is preferred to leave agnostic to let the data reveal about which channel dominates more in case of an oil price shock. No restriction is imposed on the policy rate (following other studies such as 6 Certain related studies (Corsetti, Dedola, and Leduc, 2014; Conti et al., 2017; Jovičić and Kunovac, 2017) include a variable as the share of domestic real GDP in the world output to distinguish the origin of aggregate demand shocks by imposing opposite signs. However, in order not to increase the number of variables due to the sample size being small and to avoid the over parametrization issue, this alternative approach was not preferred in this paper. 16
  20. Peersman and van Robays (2009)) as there is no optimal response of monetary authorities to an oil supply shock (like there is no optimal monetary response to domestic aggregate supply shocks). Because the oil price shocks imply both an inflationary pressure and an output contraction for the economy (Conti et al., 2017), central banks do, at least, try to not to respond and control the temporary oil price fluctuations (Bobeica and Jarocinski, 2019) until it generates a strong and persistent distorting impact on pricing behaviour in the overall economy. For these reasons, a zero restriction is put on the first-period response of US monetary policy to an oil price shock to limit an immediate reaction of the FED funds rate to a global oil price shock. A positive global demand shock is assumed to create a positive co-movement between the price of oil and global economic activity. The identification of this shock is important because it disentangles a positive domestic demand shock from that of a foreign one. Here, it is assumed that following a global demand shock, oil prices demonstrate an increase due to demand-side pressure to the commodity prices as oil is used both for consumption and production. On the other hand, a positive domestic demand shock increases both the consumer prices and the output while not affecting the oil price, global demand and FED funds rate as foreign variables are kept exogenous to domestic ones. In the estimations, grid search procedure yielded the same hyperparameters for the coefficient terms and the variance-covariance matrices. The resulting IRFs for the identified shocks in the baseline model are as follows: 17
  21. 18 Figure 3 . Baseline Model’s Impulse-Responses
  22. The domestic block , their shocks and the responses are as expected and as in the initial model’s IRF (see the Figure 2). The external block yield the following: The foreign monetary shock: Following a US monetary policy contraction of 20 bps, there is an interest rate rise in Turkey around 20 bps. Turkish lira depreciates, around 2.5 percentage points on impact. The domestic output level demonstrates a 0.5 per cent immediate decline. The explanation for that is the contraction in the US demand spills over to both advanced and emerging economies with a similar contraction in their real economies (as found by MacDonald and Popiel (2017), Canova (2005), Maćkowiak (2007), among others). This negative response can at the same time be attributed to the immediate hike in the policy rate and its contractionary impact on the real economy. The domestic price level demonstrates a significant response, around 0.10 percent of inflationary pressure. In short, the contractionary US monetary policy causes a contagion effect on interest rates of Turkey.7 When it comes to the variables in the domestic block (output, price level and exchange rate), except for the price level, it is observed that significant responses on impact with almost narrow credibility bands materialize. With regards to being agnostic, the unrestricted responses are always randomly rotated without any restrictions. Thus, this random rotation almost systematically results in IRFs being more or less flat when averaged over the draws. Therefore, not having flat IRFs in case of agnostically left variables is a significantly valuable outcome in terms of the identification. After a US monetary contraction, the global demand variable does not respond significantly as can be noticed with a flat IRF which appears as an interesting result. This can be attributed to the property that the global AD and the foreign monetary shocks are orthogonal to each other. Therefore, the foreign monetary shock is assumed to be the FED’s pure response to US demand conditions, so that the global demand to FED monetary shock does not show any respondent pattern. The crude oil price declines around 3.4 percentage points after a 20 bps hike in FFR. According to Anzuini, Lombardi, and Pagano (2012) who document the empirical relationships between the US monetary policy and the commodity prices, 7 One may argue that in the sign restriction approach, the restricted variables inherently satisfy the imposed sign. This argument is only partially true because having a significant credibility band along with the desired sign and achieving most parts of the resulting credibility band above/below the x-axis are important results to achieve in terms of the sign restriction approach in the SVAR methodology. The way the algorithm works is as follows: it generates the impulse response functions from the reduced-form VAR, implements a random rotation of those IRFs, then checks whether the conditions are satisfied. If yes, the new IRFs are retained, if not, a new attempt is made until the imposed sign restriction is achieved. Because, unless a magnitude restriction is imposed, a positive 100 per cent and a positive 1 per cent are equal in terms of the sign restrictions approach, the width of the credibility band matters. As the credibility bands are the summary information about the distribution of the total draws which are accepted according to their sign, although restricted, the response of a variable satisfying the imposed sign is crucial. 19
  23. there is a significant relationship among two . In line with the results of the seminal papers about the oil prices and the macroeconomic dynamics (Barsky and Kilian, 2004, 2001), the findings point out that US monetary policy is a significant predictor of the commodity prices. These studies reported that the impact of the monetary policy is deemed to transmit into commodity prices via the expected growth and inflation channels rather than the oil supply, oil inventories and the financial market channels. The study of Anzuini et al. (2012) finds out that 100 bps US monetary easing results in approximately 3% and 1% per cent increase in the commodity price index and oil price in particular, respectively. The SVAR results of the baseline model employed in this paper verify this inversely related correlation between the price of oil and the FED funds rate. The price of oil and the US dollar yields a similar argument as well. As the US dollar appreciates, the Brent crude oil price falls as it is priced in US dollar terms. Likewise, because the oil production rises after a US interest rates rise as the opportunity cost of keeping oil reserves on the ground rises, a fall in oil prices is an anticipated outcome. Historically, the price of oil and the price of the US dollar is inversely related. Therefore, this inverse relationship is verified by the response functions as well. Oil price shock: In case of an oil price shock with which there is a fall in the global demand, the FED does not respond initially but then it loosens its monetary policy stance gradually. The model results suggest that on impact, US dollar appreciates but after two quarters, it reverts to its original course as FED starts to lower the FED funds rate. The CBRT does not respond to a rise in the oil prices originated as a global aggregate supply shock which is in line with our expectations as the central banks do not respond to fluctuations in the commodity prices at the first place and no optimal monetary policy response exists in the standard macroeconomics literature. The consumer prices in Turkey rises following the oil price shock which makes sense considering both the exchange rate pass-through to overall prices and rising input prices (as oil is used as an intermediate input for production processes) associated with additional cost burden for the producers. The response of real output indicates a significant fall in the real GDP. It can be said that the export partners’ income rise via an increased oil price shock does not dominate the global supply concerns in the system. Rather than that, the global demand contraction channel works more predominantly. Cashin, Mohaddes, Raissi, and Raissi (2014) report that oil importers typically face a long-lived fall in economic activity in response to a supply-driven surge in oil prices, which is parallel to the results obtained at the baseline model. Global aggregate demand shock: What is left agnostic in case of a global AD shock is domestic policy rate, FED Funds rate and exchange rate. After a positive global demand shock, FED responds positively and 20
  24. significantly while there is negative but not quite a significant response from the CBRT on impact . Only after 2 to 3 quarters, a positive and significant response to the policy rate is observed. Exchange rate appreciates on impact by 1.3 per cent but then normalizes to its original path after five quarters. From the Figure 3, CBRT does not respond to the oil price movements and thus the supply-side shocks, as anticipated, but it reacts significantly to the developments in the global demand after observing the inflationary pressures are materialized. Based on the baseline model, the analyses of historical decomposition and the forecast error variance decomposition (FEVD) of the variables are presented in Appendix E. According to the FEVD results, most of the variance in the CBRT policy rate stems from the global shocks with an increasing degree as the forecast horizon increases. On the other hand, the role of the domestic aggregate demand is non-negligible. It is obvious that for the domestic output level the foreign monetary shock has more role than an oil price, global demand and a risk premium shock, while for the price level the situation is vice versa as the risk premium shocks make up most of the variation while forecasting the domestic price level. In case of the exchange rate, for all of the horizons ahead, the sum of the global shocks makes up almost the total variation. It can be inferred from the FEVD graphs that for the real macroeconomic variables, the role of the global shocks has equal weight with that of the domestic ones. However, for the nominal exchange rate and the domestic policy rate, it is clear that the global shocks dominate the domestically originated ones in the economy. When the historical decomposition plots are taken into consideration, the fluctuation in the global demand is mainly due to the global aggregate demand shocks. One can observe that the foreign monetary shocks and the oil price shocks have rather a less role in the global demand developments historically. For the FED funds rate, it is primarily driven by the US monetary shock itself as well as the risk premium shock in the global economy. Because the risk premium shock reflects all the shocks that depreciate the US dollar while lowering (increasing) the domestic price (output) level, besides the global supply, global demand and the foreign monetary shock due to the orthogonality nature of each shock, it makes sense that FED shapes its own monetary policy according to the value of the US dollar in the financial markets. When the situation in the price of oil is examined, the most dominant factor is the global real activity level. Historically, global demand developments are the dominant driving factor for oil prices. The model results support the findings of Kilian (2009) that historically, the decomposition of the fluctuations in the price of oil has been driven by a combination of global aggregate demand shocks and precautionary demand shocks for oil, rather than the supply of the oil itself. Historical decomposition of the oil prices reveals that foreign monetary shocks have a little contribution to the price of oil. 21
  25. Regarding the domestic variables , historically, domestic policy rate of Turkey is determined primarily by risk premium and foreign monetary shocks as well as by the domestic aggregate demand shock to a lesser extent. Therefore, it can be said that the CBRT also takes into account the global shocks while forming the monetary stance, apart from the domestic macroeconomic conditions. In particular, the recent developments in the bilateral exchange rate suggest that the risk premium, oil prices and the US monetary shock have played a substantial role in the course of the policy rate. As regards the price level fluctuations, the recent surge in the inflation rate is fundamentally driven by domestic demand and risk premium shocks. The oil price shocks contributed more to the consumer price fluctuations than it did to the exchange rate. The FEVD plots, on the other hand, suggests that the foreign monetary shock explains most of the variance in the USD/TL exchange rate and the policy rate of CBRT. Interestingly, the sum of all the domestic shocks makes up a little more than the global shocks while explaining the variance of the output level of Turkey. However, for the price level, policy rate and the exchange rate, the foreign shocks contribute more than the domestic shocks in all of the periods, and in particular, to an ever increasing extent as the forecast horizon increases. For the oil price, the global demand shocks determine the most of the variance, while the fluctuations of the foreign monetary shock itself and the risk premium explains the variance in the FED funds rate. 22
  26. 3 .3. A Conditional Forecasting Exercise Because the primary concern in this paper is to assess how the US monetary policy has significant implications for the Turkish economy, a conditional forecasting exercise is conducted as a scenario analysis based on the market expectations. Two consecutive monetary loosening of 25 bps at FED funds rate was anticipated during the second half of the 2019 because of the low inflationary period and the fear of a global slowdown. Following these assumptions, the Figure 4 is the conditional forecast results for the period 2019:II to 2020:IV, after imposing a 25 bps decrease in the FED funds rate for the two quarters effective for the second half of the 2019, while leaving no conditions for the rest of the shocks. In this conditional forecasting exercise, all the shocks are included in the model to generate the forecasts. By using the all shocks approach, one is certain that all of the shocks will contribute to generate the conditions, which is a plausible and informative assumption as the shocks in the model are the key structural shocks driving the domestic and the global economy. After a foreign monetary easing scenario for the second half of 2019, the results indicate a significant easing in the domestic policy rate, a fall in the consumer price level, a relatively stable exchange rate and an upward trend in the real domestic output throughout the 2020. Likewise, following this US monetary easing, a positive outlook is expected for the global production, while the oil prices are expected to rise. 23
  27. Figure 4 . Loosening FED Funds Rate Scenario 4. Robustness Check In this part by employing alternative variable choices to be used in the structural VARs and by replacing one by one the related variables in the following subsections with their alternatives, the baseline model’s data sensitivity is tested. 4.1. Alternative Shadow Rate As a robustness check, different shadow rates are tried and these alternative shadow rates also yielded the similar, robust results. Since the large portion of the estimation period corresponds to a period at which the FED engaged in unconventional monetary policy (the quantitative easing period), the choice of the shadow rates in the estimation is of vital importance. There are alternative shadow rates published by various researchers besides the shadow US federal funds rate by Wu and Xia (2016) which is adopted in this study. Below is the estimation results of the quarterly SVAR models by employing the Shadow 24
  28. US policy interest rates generated by an option-pricing model of the Krippner (2013). The identification schemes of the SVAR model is not changed for both the initial and the baseline model estimation and the estimation results are significantly robust. Figure 5. Initial Model, SVAR 25
  29. 26 Figure 6 . Baseline Model, SVAR
  30. 4 .2. Real Effective Exchange Rate Instead of using the bilateral USD/TL exchange rate, the Real Effective Exchange Rate (REER) comprising 60 countries, published by BIS Statistics Database is used as an alternative indicator to assess the relative real value of Turkish lira. The results are robust to this alternative and as follows: Figure 7. REER, SVAR 4.3. Kilian (2009) Global Real Activity Index Lutz Kilian, in one of his seminal papers (Kilian, 2009) proposes a global real activity indicator for a measure of the component of the worldwide real economic activity that drives demand for industrial commodities in global markets. His index is based on dry cargo single voyage ocean freight rates and it is explicitly designed to capture the existing shifts in demand for industrial commodities in the global markets. This is a monthly, percentage deviation from trend index and we converted it into a quarterly series by simply taking the averages. Instead of only relying on the world industrial production index (excluding the construction 27
  31. activities ) published by the CPB, I also use the Kilian’s real activity index. The impulse response results are as follows: Figure 8. Kilian Global Demand Index, SVAR 4.4. Export-Weighted Global Demand Index As another measure to control for the global demand, the export-weighted global demand index constructed specifically for Turkey is tried as well in addition to the Kilian’s global real activity index. The index is based on the study of Eren and Yavuz (2020). Exportweighted global demand index consists of 110 countries having an export coverage more than 90 percent, thus suggesting a good indicator for capturing the external demand for Turkey. The results are similar to the one yielded by the baseline model, therefore bolstering the robustness of the results. 28
  32. Figure 9 . Export-Weighted Global Demand Index, SVAR 5. Conclusion In a financially and economically integrated world, the importance of global shocks on the emerging markets has reached at a stage that the policymakers in small-open economies should consider and analyse the sources and the consequences of the external shocks. In the realm of policy rates of emerging and advanced economies move together by former following the latter, it is noteworthy to examine how the changes in global interest rates affect the macroeconomic dynamics in emerging economies. Recent literature on the transmission of the foreign monetary conditions on emerging market economies has analyzed the significance of external disturbances on these economies by adopting several methods, ranging from structural DSGE models to bilateral or panel VAR models. However, identification of foreign monetary policy shocks together with global aggregate supply and demand shocks has remained a relatively less explored research area for Turkey. As an emerging economy, Turkey appears to be an appropriate good natural laboratory to examine the global shocks as 29
  33. the domestic macroeconomic dynamics can not be considered in isolation from the external world . In this study, the identification of domestic and foreign shocks with regards to their monetary, aggregate supply and demand counterparts is attempted to quantify the impact of external shocks on Turkish economy. Our study employs a structural vector autoregressive model to the real and financial block of the Turkish economy to identify both global and domestic disturbances for the key macro-aggregates. While the estimations are computed by Bayesian procedures, the sign and zero restrictions follow the algorithm of the Arias et al. (2014), the block exogenous nature of the external shocks is from the seminal contributions to the macroeconometrics literature by the Cushman and Zha (1997). Following US monetary tightening, the results demonstrate a significant appreciation of the US dollar against the Turkish lira, a rise in overall price level and a fall in the real output level due to monetary policy reaction to the impact of exchange rate pass through to prices. On the global side, while the oil prices fall, the global demand does not manifest a significant response. Moreover, there has been a positive reaction at the policy rate in Turkey after a US monetary tightening as an anticipated policy move as a small economy. This reaction can be considered as a robust outcome of the existence of a global interest rate contagion observed in the international macroeconomics literature. Most of the research attempted to explore this contagion effect found evidence in favour of the emerging markets being a follower of the core, advanced economies. In this context, Rey (2015) argues that Mundellian trilemma could even boil down to a dilemma as independent monetary policies are only possible as long as capital flows are controlled via the macroprudential tools and measures. Apart from the analyses that the impulse response functions offer, the FEVDs presents significant information about each variable’s relative importance on forecasting a specific variable in concern. According to the FEVDs, global shocks outweigh the importance of domestic ones for predicting the financial variables of the Turkish economy, notwithstanding the domestic ones is also non-negligible. In terms of the price level and the domestic output, global and domestic shocks constitute approximately the same amount of information while explaining the variation in these variables. On the contrary, global shocks (FED monetary tightening, oil price and global demand shocks) have more role than that of domestic ones in explaining the forecast error of the financial variables. The paper contributes to the literature by investigating not only the US monetary policy shocks on the Turkish economy but also the impact of global supply and demand shocks identified by using a global demand indicator and oil prices. With regards to the connection of oil market and monetary policies, although the findings of the Kilian and Vega (2011) report no feedback effect from US macroeconomic news at 30
  34. daily and monthly frequency , the quarterly structural VAR models suggested an inverse relationship between the FED funds rate and oil prices, as supported by Anzuini et al. (2012). The results have two practical policy implications in terms of optimal macroeconomic modelling of the Turkish economy. For an empirically successful policy design, both global and domestic factors need to be taken accounted for. In this respect, advanced countries’ monetary policy choices in general or the FED’s interest rate policy path in particular matter for the monetary policy design for the domestic monetary policy. Also, as indicated by the impulse-response functions, CBRT does not respond immediately to the oil price shocks. As argued in the oil market literature, rather than the temporary developments in the oil prices, the source of the shock matters for policymakers while shaping the future monetary policy decisions. Whether it is a demand-side oil price shock that distorts the pricing behaviour and puts upward pressure on the price dynamics or a supply-side shock due to constraints of oil production is a central determinant for the central banks to take appropriate policy actions. Moreover, as the global shocks pose as much pressure as the domestic shocks to the economy, the impact of external disturbances have to be monitored closely. The paper is open to some extensions. First of all, to compare the impact of the policies of the European Central Bank(ECB) and the FED, a wider identification scheme can be utilized that encompasses the ECB policies while using European monetary policy’s shadow rates. Second, as the literature expands into a global setting incorporating countries with multiple different characteristics, a Global VAR (GVAR) model can be applied while including both ECB and FED into the system with other countries in the model. And thirdly, the timevarying parameter SVAR model can be built to account for possible historical non-linearities in the system. 31
  35. Appendix A . Algorithm For Sign Restrictions 1. Define the restriction matrices, Sj , Mj , Ml,j and Mu,j for j = 1, 2, . . . , n. 2. Define the number of successful iterations regarding the algorithm. 3. At iteration n, draw the reduced-form VAR coefficients B(n) and Γ(n) from their posterior distributions, and recover the model 1. (n) (n) (n) 4. At iteration n, obtain Ψ0 , Ψ1 , Ψ2 , . . . from B(n) . (n) (n) 5. At iteration n, calculate the preliminary structural matrix and generate Ψ0 , Ψ1 , (n) Ψ2 , . . . from Eqn. 6. Create the preliminary stacked matrix of Eqn. 12. 6. At iteration n, draw a random matrix X from a standard normal distribution. By using the QR decomposition, obtain an orthonormal matrix Q as the structural matrix. 7. At iteration n, compute a candidate structural impulse response function matrix of Eqn. 9. 8. Check if the restrictions hold. If yes, keep the matrix Q and go for the next iteration. If not, discard and repeat the steps 3 to 8, until successful number of restrictions are obtained after discarding the burnt-in iterations 32
  36. Appendix B . Technical Details The Eqn.1 can be written as an infinite order MA process as it implies the following sequence: yt = A1 yt−1 + A2 yt−2 + ... + Ap yt−p + εt (14) ⇔ yt = (A1 L + A2 L2 + ... + Ap Lp )yp + εt (15) ⇔ (I − A1 − AL2 2 . . . − Ap Lp )yt = εt (16) ⇔ A(L)yt = εt (17) where A(L) denotes the lag polinominal operator. It is possible to invert this lag polinomial as an infinite order moving average process as: A(L)yt = εt ⇔ yt = A(L)−1 εt ∞ ⇔ yt = i=0 Ψi εt-i (18) ⇔ yt = Ψ0 εt + Ψ1 εt-1 + Ψ2 εt-2 + . . . where the Ψi represents the impulse response functions of the reduced form VAR (i.e. the Eqn 1). Then, rewriting the Eqn. 18 yields: yt = BB −1 εt + Ψ1 BB −1 εt-1 + Ψ2 BB −1 εt-2 + . . . (19) which in turn implies the following by Eqn 4 at Section 2.1. yt = Bηt + (Ψ1 B)ηt-1 + (Ψ2 B)ηt-2 + . . . (20) ˜ 1 ηt-1 + Ψ ˜ 2 ηt-2 + . . . yt = Bηt + Ψ (21) and likewise as follows: or yt = ∞ i=0 ˜ i ηt-i Ψ (22) ˜ 0 ≡ B and Ψ ˜ i ’s represent the impulse response functions of the structural VAR. where the Ψ 33
  37. Appendix C . • • In the estimations, Minnesota prior is preferred. Moreover, trials with an Independent Normal Wishart prior yielded quite similar results to those of the Minnesota prior results, therefore the choice of the prior in the estimations are robust to different priors8 . The hyperparameters are chosen according to the grid search method of the Giannone, Lenza, and Primiceri (2015) that maximizes the marginal likelihood of the model. Throughout the estimations, 5000 iterations are carried out while the first 1000 of which are discarded as a burn-in sample. – – – – – 8 Model Parameters AR coefficient: 0.7 Overall tightness (λ1 ) = 0.1 Cross-variable weighting (λ2 ) = 1 Lag decay (λ3 ) = 1 Exogenous variable tightness (λ4 ) = 100 Results of different priors are available upon request 34
  38. Appendix D . Data Properties Table 3: Data Sources and Definitions Variable Source Definititon Transformation Global Variables Federal Funds Rate St. Louis FED Economic Database Quarterly averages Levels Shadow rates Wu-Xia (2013), Krippner (2013) Quarterly averages Levels Oil prices Bloomberg Brent crude oil price in US dollars QoQ-Log difference World industrial production Netherlands Bureau Policy Analysis Global real activity index Kilian(2009) for Economic World industrial production, excluding Seasonally adjusted, QoQ-Log differconstruction ence Real activity index based on dry-cargo Levels bulk freight rates Domestic Variables Output TURKSTAT, CBRT Real production level Seasonally adjusted, QoQ-Log difference Price level TURKSTAT Consumer price index level Seasonally adjusted, QoQ-Log difference Policy rate CBRT, BIST Main policy interest rate levels USD/TRY Exchange Bloomberg rate Nominal exchange rate against USD Monthly averages dollar REER Real effective exchange rate of Turkey BIS Statistics 35 QoQ-Log differences
  39. Table 4 : Unit Root Tests ADF Test KPSS Test H0 = Unit root H0 = Stationary Level 1st Difference Level 1st Difference Shadow Rates -2.050 -2.327 .400 .172 CBRT Policy Rate -1.767 -4.852 .585 .683 TR/USD Exchange Rate -6.843 -8.919 .779 .194 World Industrial Production -1.438 -3.495 .983 .044 Oil Price -2.660 -6.301 .254 .156 Domestic Production -0.205 -5.619 1.006 .085 Consumer Price Level 3.670 2.179 1.013 .802 Critical Values %1 -3.553 .216 %5 -2.915 .146 %10 -2.592 .119 36
  40. Figure 10 . International Policy Interest Rates∗ * Median across Chili, Indonesia, Russia, Hungary, Mexico, India, South Africa, Brazil, Poland, Malaysia, Peru, Columbia, Romania and Turkey, where data are available. Source: BIS Statistics, Wu-Xia(2015) Figure 11. Effective FED funds rate and CBRT Policy Rates Source: FRED Economic Data, CBRT, Bloomberg 37
  41. Appendix E . Variance Decomposition and Historical Decomposition Analysis Table 5: Price Level, Variance Decomposition Horizon 2 4 6 8 12 Domestic monetary 0.08 0.08 0.08 0.08 0.08 Horizon 2 4 6 8 12 Domestic monetary 0.08 0.08 0.08 0.08 0.08 Horizon 2 4 6 8 12 Domestic monetary 0.06 0.05 0.05 0.05 0.05 Domestic Supply Domestic Demand Risk Premium Global Demand 0.35 0.13 0.15 0.03 0.33 0.12 0.15 0.04 0.32 0.12 0.15 0.05 0.32 0.12 0.15 0.05 0.31 0.12 0.15 0.05 Global Supply (Oil Price) 0.06 0.07 0.07 0.07 0.07 Foreign monetary 0.02 0.02 0.03 0.03 0.03 Table 6: Domestic Output Level, Variance Decomposition Domestic Supply Domestic Demand Risk Premium Global Demand 0.25 0.09 0.05 0.07 0.24 0.09 0.05 0.08 0.24 0.09 0.05 0.08 0.24 0.09 0.06 0.08 0.24 0.09 0.06 0.08 Global Supply (Oil Price) 0.12 0.12 0.12 0.12 0.12 Foreign monetary 0.10 0.10 0.10 0.10 0.10 Table 7: CBRT Policy Rate, Variance Decomposition Domestic Supply Domestic Demand Risk Premium Global Demand 0.05 0.27 0.34 0.01 0.05 0.25 0.31 0.03 0.05 0.24 0.28 0.04 0.05 0.23 0.26 0.05 0.05 0.22 0.24 0.06 38 Global Supply (Oil Price) 0.004 0.006 0.008 0.01 0.01 Foreign monetary 0.07 0.12 0.13 0.15 0.17
  42. Figure 12 . Historical Decomposition of the Model Variables CPI GDP STN 8 4 3 3 2 6 1 4 2 0 2 1 -1 0 -2 0 -2 -3 -4 -4 -1 -6 -5 -2 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 E 2010 2012 2014 2016 2004 2018 2006 2008 2010 GlobalDemand 2012 2014 2016 2018 Oil 20 20 1 10 15 0 10 -1 -10 -2 -20 0 5 -30 -3 -40 0 -4 -5 -50 -5 -60 -6 -10 2004 2006 2008 2010 2012 2014 2016 2018 2004 2004 2006 2008 2010 2012 2014 Shadow 3 Domestic monetary 2 Domestic AS 1 Domestic AD 0 Risk Premium -1 Global AD Shock -2 Oil Price (Global AS) Shock -3 2004 2006 2008 2010 2012 2014 2016 2018 Foreign monetary 39 2016 2018 2006 2008 2010 2012 2014 2016 2018
  43. Figure 13 . Forecast Error Variance Decomposition of the Model Variables CPI 100 GDP 100 80 80 80 60 60 60 40 40 40 20 20 20 0 0 1 2 3 4 5 6 7 8 9 10 11 12 E 100 0 1 2 3 4 5 6 7 8 9 10 11 12 GlobalDemand 100 1 80 80 60 60 60 40 40 40 20 20 20 0 0 2 3 4 5 6 7 8 9 10 11 12 2 3 4 5 6 7 8 Domestic monetary 80 Domestic AS 60 Domestic AD Risk Premium 40 Global AD Shock 20 Oil Price (Global AS) Shock Foreign monetary 0 1 2 3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11 12 7 8 9 10 11 12 Oil 0 1 Shadow 100 2 100 80 1 STN 100 12 40 9 10 11 12 1 2 3 4 5 6
  44. References Anzuini , A., Lombardi, M. J., Pagano, P., 2012. The impact of monetary policy shocks on commodity prices. Bank of Italy Temi di Discussione Working Paper . Arias, J. E., Rubio-Ramirez, J. F., Waggoner, D. F., 2014. Inference based on svars identified with sign and zero restrictions: Theory and applications. Dynare Working Papers 30. Barsky, R. B., Kilian, L., 2001. Do we really know that oil caused the great stagflation? a monetary alternative. NBER Macroeconomics annual 16, 137–183. Barsky, R. B., Kilian, L., 2004. Oil and the macroeconomy since the 1970s. Journal of Economic Perspectives 18, 115–134. Bobeica, E., Jarocinski, M., 2019. Missing disinflation and missing inflation: A var perspective. International Journal of Central Banking 15, 199–232. Buyukbasaran, T., Can, G. K., Kucuk, H., 2019. Identifying credit supply shocks in turkey. CBRT Working Paper Series . Caceres, C., Carriere-Swallow, Y., Demir, I., Gruss, B., 2016. U.s. monetary policy normalization and global interest rates. IMF Working Papers . Canova, F., 2005. The transmission of us shocks to latin america. Journal of Applied econometrics 20, 229–251. Cashin, P., Mohaddes, K., Raissi, M., Raissi, M., 2014. The differential effects of oil demand and supply shocks on the global economy. Energy Economics 44, 113–134. Conti, A. M., 2017. Has the fed fallen behind the curve? evidence from var models. Economics Letters 159, 164–168. Conti, A. M., Neri, S., Nobili, A., 2017. Low inflation and monetary policy in the euro area. ECB Working Paper Series . Corsetti, G., Dedola, L., Leduc, S., 2014. The international dimension of productivity and demand shocks in the us economy. Journal of the European Economic Association 12, 153–176. Cushman, D. O., Zha, T., 1997. Identifying monetary policy in a small open economy under flexible exchange rates. Journal of Monetary economics 39, 433–448. 41
  45. Dedola , L., Rivolta, G., Stracca, L., 2017. If the fed sneezes, who catches a cold? Journal of International Economics 108, S23–S41. Demir, I., 2019. International spillovers of u.s. monetary policy. Lincoln Economics and Finance Research Group (LEAF) Working Paper Series . Dieppe, A., Van Roye, B., Legrand, R., 2016. The bear toolbox. ECB Working Papers . Dungey, M., Pagan, A., 2009. Extending a svar model of the australian economy. Economic Record 85, 1–20. Eren, O., Yavuz, D., 2020. Bölgeler İtibarıyla İhracat ağırlıklı büyüme endeksleri. CBRT Research Notes . Georgiadis, G., 2016. Determinants of global spillovers from us monetary policy. Journal of International Money and Finance 67, 41–61. Gertler, M., Karadi, P., 2015. Monetary policy surprises, credit costs, and economic activity. American Economic Journal: Macroeconomics 7, 44–76. Giannone, D., Lenza, M., Primiceri, G. E., 2015. Prior selection for vector autoregressions. Review of Economics and Statistics 27, 436–451. Gupta, P., Masetti, O., Rosenblatt, D., 2017. Should emerging markets worry about US monetary policy announcements? The World Bank. Hajek, J., Horvath, R., 2018. International spillovers of (un) conventional monetary policy: The effect of the ecb and the us fed on non-euro eu countries. Economic Systems 42, 91–105. Hofmann, B., Takáts, E., 2015. International monetary spillovers. BIS Quarterly Review . Jovičić, G., Kunovac, D., 2017. What is driving inflation and gdp in a small european economy: The case of croatia. Croatian National Bank . Kilian, L., 2009. Not all oil price shocks are alike: Disentangling demand and supply shocks in the crude oil market. American Economic Review 99, 1053–69. Kilian, L., Vega, C., 2011. Do energy prices respond to us macroeconomic news? a test of the hypothesis of predetermined energy prices. Review of Economics and Statistics 93, 660–671. 42
  46. Krippner , L., 2013. A tractable framework for zero-lower-bound gaussian term structure models . MacDonald, M., Popiel, M. K., 2017. Unconventional monetary policy in a small open economy. International Monetary Fund. Maćkowiak, B., 2007. External shocks, us monetary policy and macroeconomic fluctuations in emerging markets. Journal of Monetary Economics 54, 2512–2520. Peersman, G., van Robays, I., 2009. Oil and the euro area economy. Economic Policy 24, 603–651. Rey, H., 2015. Dilemma not trilemma: The global financial cycle and monetary policy independence. CEPR Discussion Papers . Smets, F., Wouters, R., 2003. An estimated dynamic stochastic general equilibrium model of the euro area. Journal of the European Economic Association 1, 1123–1175. Szafranek, K., Hałka, A., et al., 2017. Determinants of low inflation in an emerging, small open economy. a comparison of aggregated and disaggregated approaches. Tech. rep., Narodowy Bank Polski, Economic Research Department. Uhlig, H., 2005. What are the effects of monetary policy on output? results from an agnostic identification procedure. Journal of Monetary Economics 52, 381–419. Wu, J. C., Xia, F. D., 2016. Measuring the macroeconomic impact of monetary policy at the zero lower bound. Journal of Money, Credit and Banking pp. 253–291. 43
  47. Central Bank of the Republic of Turkey Recent Working Papers The complete list of Working Paper series can be found at Bank ’s website (http://www.tcmb.gov.tr) Okun’s Law under the Demographic Dynamics of the Turkish Labor Market (Evren Erdoğan Coşar, Ayşe Arzu Yavuz Working Paper No. 21/08, March 2021) Potential Growth in Turkey: Sources and Trends (Orhun Sevinç, Ufuk Demiroğlu, Emre Çakır, E. Meltem Baştan Working Paper No. 21/07, March 2021) Cost of Credit and House Prices (Yusuf Emre Akgündüz, H. Özlem Dursun-de Neef, Yavuz Selim Hacıhasanoğlu, Fatih Yılmaz Working Paper No. 21/06, March 2021) External Vulnerabilities and Exchange Rate Pass-Through: The Case of Emerging Markets (Abdullah Kazdal, Muhammed Hasan Yılmaz Working Paper No. 21/05, February 2021) The Impact of Oil Price Shocks on Turkish Sovereign Yield Curve (Oğuzhan Çepni, Selçuk Gül, Muhammed Hasan Yılmaz, Brian Lucey Working Paper No. 21/04, February 2021) Decomposition of Bank Loans and Economic Activity in Turkey (Hande Küçük Yeşil, Pınar Özlü, Çağlar Yüncüler Working Paper No. 21/03, February 2021) The Role of Expectations in the Inflation Process in Turkey: Have the Dynamics Changed Recently? (Ümit Koç, Fethi Öğünç, Mustafa Utku Özmen Working Paper No. 21/02, February 2021) Consequences of a Massive Refugee Influx on Firm Performance and Market Structure (Yusuf Emre Akgündüz, Yusuf Kenan Bağır, Seyit Mümin Cılasun, Murat Günay Kırdar Working Paper No. 21/01, January 2021) Do Household Consumption and Saving Preferences Vary Between Birth-Year Cohorts in Turkey? (Evren Ceritoğlu Working Paper No. 20/15, October 2020) Credit Decomposition and Economic Activity in Turkey: A Wavelet-Based Approach (Oğuzhan Çepni, Yavuz Selim Hacıhasanoğlu, Muhammed Hasan Yılmaz Working Paper No. 20/14, October 2020) Do Investment Incentives Promote Regional Growth and Income Convergence in Turkey? (Hülya Saygılı Working Paper No. 20/13, October 2020) An Analysis of International Shock Transmission: A Multi-level Factor Augmented TVP GVAR Approach (Bahar Sungurtekin Hallam Working Paper No. 20/12, October 2020) Synchronization, Concordance and Similarity between Business and Credit Cycles: Evidence from Turkish Banking Sector (Mehmet Selman Çolak, Abdullah Kazdal, Muhammed Hasan Yılmaz Working Paper No. 20/11, October 2020) Identification of Wealthy Households from the Residential Property Price Index Database for Sample Selection for Household Surveys (Evren Ceritoğlu, Özlem Sevinç Working Paper No. 20/10, October 2020) Corporate Debt Maturity, Repayment Structure and Monetary Policy Transmission (Hatice Gökçe Karasoy Can Working Paper No. 20/09, May 2020) Detecting the Position of Countries in Global Value Chains Using a Bilateral Approach (Oğuzhan Erdoğan Working Paper No. 20/08, May 2020)