IslamicMarkets uses cookies.
About our cookie policy.
  of  

or
Sign in to continue reading...

Composing High-Frequency Financial Conditions Index and Implications for Economic Activity

Abdullah Kazdal
By Abdullah Kazdal
1 week ago
Composing High-Frequency Financial Conditions Index and Implications for Economic Activity

Create FREE account or Login to add your comment
0 Comments


Transcription

  1. Composing High-Frequency Financial Conditions Index and Implications for Economic Activity Abdullah KAZDAL Halil İbrahim KORKMAZ Muhammed Hasan YILMAZ September 2019 Working Paper No: 19/26
  2. © Central Bank of the Republic of Turkey 2019 Address: Central Bank of the Republic of Turkey Head Office Structural Economic Research Department Hacı Bayram Mh. 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. The Working Paper Series are externally refereed.
  3. Composing High-Frequency Financial Conditions Index and Implications for Economic Activity Abdullah Kazdal Halil İbrahim Korkmaz Muhammed Hasan Yılmaz1 Abstract In this study, the main aim is to construct an index using high-frequency data related to financial markets and intermediation services for Turkey, termed as High-Frequency Financial Conditions Index, by employing alternative statistical techniques. In a complementary manner, the informative nature of the constructed indices with respect to the course of economic activity is examined. The paper also includes detailed empirical analysis about the relationship between financial conditions and growth tendencies. The findings of the time series analysis and forecast exercises show that the constructed series are quite informative regarding the economic activity. More importantly, probit model estimations indicate that index can be qualified as an early indicator to predict “loss of momentum” episodes in economic growth by also considering the lead-lag relationship. When similar methodology is applied on emerging market economies, indices can be produced with a high level of co-movements with growth indicators. Panel Vector Autoregression estimation shows that, after controlling for country-specific characteristics, a shock coming to financial conditions is creating a significant overall response in emerging market countries. In terms of policy-making, we believe that constructed indices will contribute to a better understanding of the current financial environment and relation with economic activity. Özet Bu çalışmada alternatif istatistiksel teknikler yardımıyla Türkiye için finansal piyasalar ve aracılık hizmetleri ile ilgili yüksek frekanslı veriler kullanılarak Yüksek Frekanslı Finansal Koşullar Endeksi oluşturulması amaçlanmaktadır. Ayrıca, oluşturulan endeksin iktisadi faaliyete ilişkin açıklayıcı niteliği de incelenmiştir. Ek olarak, finansal koşullar ile büyüme eğilimleri arasındaki ilişki yüksek frekanslı verilerle ampirik olarak analiz edilmektedir. Çalışmanın bulguları, oluşturulan serinin ekonomik faaliyet üzerindeki açıklayıcı gücünün oldukça yüksek olduğunu göstermektedir. Bunun yanında, türetilen endeks ile çıktı göstergeleri arasındaki gecikmeli ilişki dikkate alındığında, endeksin büyümede ivme kayıpları yaşanan dönemler açısından bir öncü gösterge niteliği taşıdığı anlaşılmaktadır. Benzer metodoloji gelişmekte olan ülke ekonomileri için uygulandığında, bu ülkeler için de büyüme göstergeleriyle beraber hareket eden endeksler türetilebilmektedir. Ülke bazlı özellikleri kontrol ederek gerçekleştirilen panel vektör özbağlanım tahminleri, gelişmekte olan ülkelerde finansal koşullara gelen şokların iktisadi faaliyet üzerinde toplu olarak anlamlı etki oluşturduğuna işaret etmektedir. Politika yapıcılar açısından, oluşturulan endeksin, finansal gelişmelerin ekonomik faaliyetle ilişkisinin daha iyi anlaşılmasına katkıda bulunacağına ve zamanlı politika adımlarının atılmasına katkı sağlayacağı düşünülmektedir. JEL Classification: G10, E17, E44, E66. Keyword: Financial Conditions, Growth Dynamics, Factor Models, Forecasting, Probit Models, Panel VAR. 1Central Bank of the Republic of Turkey. E-mail adresses: halil.korkmaz@tcmb.gov.tr, abdullah.kazdal@tcmb.gov.tr, muhammed.yilmaz@tcmb.gov.tr. The views expressed in this paper are those of the author and do not reflect the official views of the Central Bank of the Republic of Turkey. We gratefully acknowledge comments by İbrahim Ethem Güney, Oğuzhan Çepni, Mehmet Selman Çolak and Doruk Küçüksaraç. We also want to thank anonymous referee for useful suggestions. 1
  4. Non-Technical Summary Given the occurrence of recent financially volatile episodes , one of the most important focal points of policymaking has re-emerged as monitoring financial developments. Furthermore, drastic changes in financial outlook have been empirically associated with economic downturns, especially in the case of Emerging Market (EM) economies with considerable leverage in balance sheet positions and profound dependence on external financing. In addition to this complex interconnectedness, the rapid increase in the amount of available data makes analysis much harder for economists who are monitoring the linkages between financial and real variables. Thus, compact indicators summarizing the current financial situation and giving some clues about the future course of the economy are thought to provide convenience in terms of taking timely and appropriate proactive policy measures. Financial Condition Index (FCI) is the standard format of such indicators representing the tightness/looseness of the financial environment via summarizing a group of financial indicators. The FCIs in the existing literature differ in terms of the scope (included variables), frequency (quarterly, monthly or weekly) and statistical methodology. In this study, our aim is to construct an FCI using high-frequency data related to financial markets and intermediation services regarding Turkey. Another goal of this paper is to examine the informative nature of constructed High-Frequency Financial Conditions Index (HFFCI) for the course of economic activity. In this perspective, the study also includes empirical analysis about the relationship between financial conditions and growth tendencies. We initially provide descriptive analysis regarding the degree of co-movement between economic activity and HFFCI. In the following phase, the predictive power of HFFCI for growth tendencies are investigated with a forecasting exercise. Lastly, the role of HFFCI in predicting the probability of loss of momentum in growth is assessed. On the back of the profound relationship with economic activity and important predictive power for Turkey, we extend the similar methodology to EM economies. This extension yields indicators successfully capturing the financial outlook in those countries. When historical relations are examined, it is observed that there are some financially volatile episodes during which EM economies have experienced a common tightening in terms of financial conditions and they have faced with a coincided weakening in economic growth tendencies. The analysis made with panel data of EM countries also validates the existence of such a common behavior that can be defined as the response of economic growth to shocks coming to local financial conditions. 2
  5. 1 . Introduction and Related Literature After the Global Financial Crisis (GFC), financial stability concerns have emerged with the awareness of more globalized, interconnected and complicated financial markets. In addition to complexity, the rapid increase in the amount of available data makes analysis much harder for economists who are monitoring the linkages between financial and real variables. Therefore, compact indicators summarizing the current financial situation and giving some clues about the future course of the economy have become popular and useful. Such indicators provide convenience in terms of monitoring the recent period and flexibility for taking timely and appropriate proactive policy measures. Financial Condition Index (FCI) is the standard format of such indices representing the tightness/looseness of the financial environment via summarizing a group of financial indicators. In addition to this, considering the monetary policy transmission mechanism and lead-lag relations, FCI might have important implications for output and price level of the country. Although the pioneering studies regarding FCI concentrated on developed markets (mainly the US economy), there is a growing body of literature focusing on Emerging Markets (EM). In terms of studies focusing on advanced economies, Swiston (2008) attempt to construct FCI for the US by employing vector autoregressive (VAR) model and impulse response function (IRF) analysis to determine the weights of the sub-components of the index. Guichard and Turner (2008) also follow similar methodology (VAR and reduced-form equations) to come up with FCI for the US by using variables such as exchange rate, interest rates, bond spread, and some asset prices. On the other hand, Hatzius et al. (2010) use the principal component analysis (PCA) method to obtain FCI and examine the predictive performance of the index for economic activity. Regarding the Euro Area, Montagnoli and Napolitano (2005) use Kalman-Filter methodology to obtain the weights assigned to each variables constituting the FCI of the Euro Area and US. Besides that, Angelopoulou (2014) benefit from the PCA method to build FCI for some European countries. Considering the studies covering EMs, Gomez (2011) construct FCI for Colombia by adapting PCA methodology on a broad range of variables comprising interest rates, exchange rates and asset prices. Cottani et al. (2012) build an indicator to summarize the state of financial conditions in Latin American countries. Apart from that, Osorio et al. (2011), construct a 3
  6. quarterly FCI for 13 Asian economies including the developing ones . The authors create FCIs based on two main statistical techniques which are VAR model and dynamic factor model (DFM). Gumata et al. (2012) work on an FCI for South Africa that is based on both global and domestic financial indicators via combining PCA and DFM. For Turkey, there are some studies focusing on FCI construction. As influential and pioneering works, Kara et al. (2012, 2015) build quarterly FCI series for the Turkish economy by employing VAR methodology while using selected variables based on expert judgment and various methods. Then, the predictive performance of the constructed FCI series for output growth is examined. In particular, Kara et al. (2015) identify a broader set of variables embodying financial market-based information about the exchange rate market, equity market, risk premia and bond markets. They also include variables representing capital flows, banking sector outlook, housing prices and money supply. After the identification of this broad list, they embrace a subjective approach and try to form a smaller sub-group by using the expert judgment. However, in order to make more robust inferences, they aim to include variables with longer time series and to cover all of the financial markets. This approach is further supported with econometric tests to analyze the informative nature of excluded variables in the first step. In particular, they run linear regressions of GDP growth on constructed FCI and excluded variables with four lags. Then, they conduct joint F-tests to identify whether or not excluded candidate variables’ lags are jointly insignificant or not. In the following stage, impulse-responses generated from VAR models including growth tendencies and financial variables are taken as inputs in determining the appropriate weights of each individual variable in ultimate FCI indexation procedure. As it becomes evident with the earlier studies in the literature, The FCIs differ in terms of the scope (included variables), frequency (quarterly, monthly or weekly) and statistical methodology (PCA, DFM or VAR). In this study, our aim is to construct an FCI using highfrequency data related to financial markets and intermediation services. Another goal of this paper is to examine the informative nature of constructed FCI for the course of economic activity. In this perspective, the study also includes empirical analysis about the relationship between financial conditions and growth tendencies with high-frequency data. 4
  7. Section 2 introduces the framework for estimating and composing High-Frequency Financial Conditions Index (HFFCI) by providing detailed information about PCA and DFM methodologies. Section 3 presents empirical results regarding the predictive power of HFFCI for economic growth in Turkey through time series analysis, forecasting exercises as well as probit models. Section 4 applies the same methodology of index construction to selected EM economies and presents the results of panel VAR estimations regarding the association between local financial conditions and economic growth by also taking countryspecific fixed effects into consideration. The last section concludes the discussion in the paper. 2. Methodology 2.1. Composition and Formation of HFFCI Generally, HFFCI-type indices are obtained through PCA, DFM and VAR models in line with the literature and the previous studies for Turkey. Each method enables one to exploit different advantages but one important issue to be clarified is the selection of data. In this study, we aim to provide an HFFCI with a clear economic intuition. In other words, instead of choosing statistically best-performing components though variable selection procedures, we choose specific variables that are easy to follow under the scope of monetary transmission mechanism2. This includes equity, bond, portfolio flows, exchange rate, and credit markets together with market inflation expectations and attached credit risk. In this sense, the traditional interest rate channel, exchange rate channel and asset prices channel of monetary transmission mechanism among others are thought to be represented by our variable set. Table 1 provides descriptive information for each indicator used for the construction of HFFCI. All variables are collected for the period covering after 2005, but because of the data conversion issues, the HFFCI is formed from the June 2006 till December 2018 on a weekly frequency through PCA and DFM. HFFCIs constructed through PCA and DFM basically summarize common movements in chosen variables, thus resulting index serves as a composite unobservable indicator of the financial conjuncture. Unlike common factor 2 TCMB, Parasal Aktarım Mekanizması, 2013, www.tcmb.gov.tr ISBN (elektronik): 978-605-5758-89-9 5
  8. models, VAR analysis requires a priori model specification where the chosen variables are initially regressed on a growth variable, and then, the cumulative coefficients are obtained from an impulse-response function to specify the weights of individual data within the indexation mechanism. However, since VAR analysis requires the judgmental choices in variable selection due to existence of highly correlated variables in the base date set and model specification processes in addition to the observation loss caused by the lower frequency of GDP growth or IPI series in extracting weights, we choose to proceed with factor models to obtain the HFFCI on a weekly basis. Nonetheless, as a robustness check, similar data is also aggregated with VAR model. The graphical investigation provided in the Appendix points out that indices generated with basic static factor model and VAR do not deviate from each other in an influential manner. Each variable in Table 1 is available in public data sources whereas only yield curve variables are obtained through a prior yield curve fitting. Yield curve is estimated through Nelson and Siegel (1987) methodology for 1m-10y maturity cross currency swap rates. Nelson-Siegel (NS) model is one of the most commonly used parametric yield curve estimation methods due to reliable interpretation of its coefficients as level, slope and curvature. Spot rate function