Arnoldshain Seminar XI. Migration, Development, and Demographic Change – Problems, Consequences, Solutions June 25 – 28, 2013, University of Antwerp, Belgium Argentine Terms Of Trade Volatility Handling Structural Breaks And Expectation Errors José Luis Arrufat Alberto M. Díaz Cafferata Santiago Gastelú Instituto de Economía y Finanzas. Facultad de Ciencias Económicas Universidad Nacional de Córdoba
Introduction • Literature review • Breaks in Argentine terms of trade and GDP • Approaches to measuring volatility • Empirical estimation of GDP and TOT volatility • Exploratory analysis of causality • Concluding remarks
I Introduction Currentprominence of volatility in developmenteconomics: impactongrowth. Whatisvolatility? Howhigh? Howdoesitbehavealong time?
Argentina TOT 1810-2010. Index 1993=100. Large & suddenchanges. Extreme peaks and valleysFourstructuralbreaks 1882, 1913, 1945, 1975. * Structural breaks 1909: 146 1948: 150 1922: 71 1987: 85 2000: 106 2010: 141 1950 * 1839 * 1917 *
Argentina TOT index, 1810-2012 Highobservedfluctuations. Mean = 97.05; SD = 22.46; CV = 0.23 High TOT volatilityis a characteristic of commodity-exporterdevelopingcountries. TOT volatility developing countries, 3 times higher than industrial countries. (Aizenmanet al. 2011, Mendoza 1995) Doesitmatter?
SOE “vulnerability” toexternal shocks and volatility. Do TOT matter? Theanswer, two temporal frameworks. Macroeconomicperspective o-f-all unexpected TOT shocks → “cause” CA shifts? Harberger-Laursen-Metzlereffect. Sign of transitoryorpermanent, shocks. Models w/woinvestment. Long-termdevelopment Effects of uncertainty: volatility onrate & volatility of growth, distribution and poverty.
Shocks & macro Literatureonthe HLM effect. “TOT matter” Harberger, Arnold C. 1950 "Currency Depreciation, Income, and the Balance of Trade." JPE . (58). Laursen Sven & Metzler Lloyd A., 1950 “Flexible Exchange Rates and the Theory of Employment.” Review of Ec & Statistics, (32) 3 . Obstfeld Maurice, 1982 "Aggregate Spending and the Terms of Trade: Is There a Laursen-Metzler Effect?“ Quarterly J Economics (97) 2. 1950. The “HLM effect”: positive relationship between TOT shocks and the CA. Income rises and C rises less. 1981. Obstfeld, Sachs, Svenson & Razin: the CA improves only if the TOT shock is transitory (otherwise there is not a smoothing role for the CA) 1990. Mendoza & Otto: there is an HLM effect with both transitory and permanent shocks)
Barone Sergio V. , Ricardo L. Descalzi, Alberto M. Díaz Cafferata (2009) “Terms of Trade Shocks and Current Account Adjustment”. XXIV JornadasAnuales de Economía. BCU 18 LACs, in 1976-2007. Data: BM y FMI. Model FGLS TOT matter for the CA: Estimated coefficient for the permanent TOT shock significantly different from zero, and positive sign.
Ourfocus: volatility & growth • Perceived costs of high and irregular fluctuations along time. • Attentionshiftsfrom SR impact of shockstowardseffects of volatilityongrowth Problem: shared intuition, but not an agreed empirical measure of TOT volatility in quantitative estimations,
Methodologicalissues To quantify magnitude, and effects What is formally “volatility”? How high? It depends on how you measure it.
Volatilityisnotaninequivocalconcept Several definitions depict different temporal profiles! How do different methods compare? What criteria to choose to depict stylized facts and estimate association? Compare below patterns with three methods
FIGURE V.1. TOT VOLATILITYDetrendedcumbreaks (BLUE) Detrened cum breaks + decycled (RED, lowervolatility!) * Sample: 1840 to 2012. 30-year rolling sample SD. * 1839 * 1917 * 1951
FIGURE V.3 IS VOLATILITY REALLY OVERESTIMATED? (Friedman-Cavallo)
II Literaturereview. Empiricalestimation of volatility and structuralbreaks
Magnitude and impacts of volatility Broadrange of topics Howhighisvolatility (empiricalestimation), Measureuncertainty (methodstoportray) Causes (specialization & markets) Channels and effectson GDP growth and distribution Weaknesses of developingcountries. Policyrecommendations
“Prominence of volatility” Aizenman and Pinto 2005, p2. Volatility has a central place in development economics. What has catapulted volatility into this prominence? Negative impacts on trend growth, Effects on saving & investment, and links between technological progress and the capital stock Understanding the nature of volatility, anticipating and managing its consequences, is of considerable interest to policymakers in developing countries.
Volatilitymattersforgrowth Mendoza 1997 “TOT are typically a significant and robust determinant of economic growth”. Model savings under uncertainty. Aizenman and Pinto (2005) large growth cost especially for developing countries. Wolf (2005) a growing body of research suggests that higher volatility is causally associated with lower growth. Loayza and Raddatz (2007) 25% of the variation in growth volatility. Koren and Tenreyro (2007)
WOW HIGH IS TOT VOLATILITY?Howdoesitevolves? Ourfocus, tackle EMPIRICAL ESTIMATION OF VOLATILITY Note it INVOLVES METHODOLOGICAL ISSUES Weadoptan EXPECTATIONS-BASED PERSPECTIVE
Metodologicalissuesin theempiricalestimation of volatility Critique to the purely statistical approach: distinguish volatility from variability (Dehn, Wolf) Filter perceived trend (problem: choice of detrending method; Canova, Bee de Dagum) Time varying volatility (use of rolling window; Ramey and Ramey; Arrufat et al)) Large jumps vs smooth trends (finding breaks; Ocampo & Parra, Bai-Perron) Filter perceived regular cycles (Bolch & Huang; determine cycles included) Deal with temporal anachronism (agent´s dataset and knowledge of DGP; Cavallo, Friedman)
(a) Empirical estimation of volatility: the statistical approach Perry (2009) SD of cyclicalcomponentfromthetrend Aizenmanet al. (2011), “Adjustment patterns to commodity terms of trade shocks: the role of exchange rate and international reserves policies”, NBER WP 17692. Larrain & Parro (2006), “Chile menosvolátil”, Instituto de Economía, U. Católica de Chile. This method depict observed fluctuations. Does it measure volatility?
(b)Expectations-based volatility Decompose observed data on predictable (regular part) and unpredictable (uncertainty) components. • Kim (2007)“Openness, external risk, & volatility: implications for the compensation hypothesis”, Cambridge UP • Wolf “Volatility: Definitions and Consequences”, In Aizenman & Pinto Managing Volatility and Crises. • Dehn (2000), "Commodity price uncertainty in developing countries”, World Bank (Series 2426) • Baxter (2000), “International trade and business cycles”, in Grossman and Rogoff .
Abruptchanges in TOT. Severalauthors note thepresence of breaks • Ocampo and Parra-Lancourt (2010b) barter TOT for commodities vs manufactures improved declined since the early 20th century with a stepwise deterioration in 1920 and 1979. • Cuddington and Urzúa (1989) the real commodity price index drops abruptly in 1921; there is no evidence of an ongoing secular deterioration. • Bleaney & Greenaway (1993)
Reasonstoidentifybreaks. Empirical research has found significant episodes of large jumps in TOT. Portraying stylized facts. Improve analysis identifying changes in DGP and structural differences in regimes between breaks. Detrending method in the presence of breaks, Are there breaks in Argentine TOT and GDP?
III Structural breaks in Argentina. TOT and GDP First step in the estimations. Breakpoints: Bai-Perron test. Different regimes.
Reasonsto test forbreaks Avoid erroneous characterizations of the nature of the series. (e.g. mistakenly arriving at the conclusion that a series is stationary in differences when it is in fact trend stationary but with a segmented trend). Severe pitfalls may arise in the process to isolate cycles. An important outlying observation may lead the researcher to identify a bogus cycle the period of which is excessively lengthy.
Bai – Perron test formbreaks A segmented trend for the first subperiod :T0 to T1-1 If there is one break, the second subperiod runs between T1 and T2-1 … If there are m breaks, the expression for the m+1 regime is: All summations run from 0 to m
ESTIMATION OF TOT BREAKS Noticethatbreaksoccur in 1839, 1917, and 1951. Dummiesthatwerenotsignificantlydifferentfromzeroweredroppedtoensurethemostparsimoniousmodel.
CAMBIAR LA FILMINA POR OTRA • CON ERRORES ESTÁNDAR ROBUSTOS • DADA LA PRESENCIA DE ALTA AUTO CORRELACIÓN
TheBayesianInformationCriterion • There is a trade-off between goodness of fit (the residual sum of squares RSS) measured on the right axis , which is monotonically non- increasing with the number of breaks, and parsimony. • The Bayesian Information Criterion (BIC) takes into account both goodness of fit and parsimony. • The minimum BIC in the TOT is for three breaks. There are four breaks in the GDP series.
Argentina 1810-2012. TOT Break-points Four TOT regimes. Break years: 1839, 1917, 1951
ESTIMATION OF GDP BREAKS Notice that breaks occur in 1882, 1913, 1945, and 1975. Dummies not significantly different from zero were dropped to ensure a parsimonious model.
Argentina 1810-2012. GDP Break-points Five GDP regimes. Break years: 1882, 1913, 1945, 1975
Dating of breakpoints for logTOT and logGDP, and other sources of epochs The Baring Crisis was in 1890. Cfr. Cortés Conde, la economía argentina en el largo plazo. Díaz Cafferata “Inercia estructural del crecimiento”: Academia Nac Cs Económicas, after Max trend growth decline secularly with trade openness until the 1980s
Detrending cum breaks • Both TOT and GDP exhibit breaks that shall be taken into account in the decycling. • The break-points point out a transformation or transition zones. • Years of breaks estimated make sense: portray three great economic history epochs of Argentine: first one the open, golden XIXth Century high growth, like other land abundant countries, until the first World War (Baring crisis 1890) with four decades of transition between 1875 and 2014. A second one is the interwar period of relatively low trade openness. The third one the last half-century of globalization.
IV Measuring volatility with alternative methods. A discussion.
How much “volatility”? Volatility analytical interpretation: associated with uncertainty. Proxy in standard empirical practice, through two approaches.
Measuring volatility. Our taxonomy of approaches to volatility. Different definitions of volatility in the literature, can be grouped in two main empirical approaches (a) Statistical SD of a time series SD of detrended residuals (b)Expectations-based b.1. Detrending + Decycling b.2. Forecasting errors
(a) Statistical approach Original Series. Statisticalapproach. Descriptive measures of dispersion. SD of a time series SD of detrended residuals. Single value or rolling sample. Volatility measured by the SD: may be a single global value of the period, or a rolling window which provides a temporal profile. Measures fluctuations of observed series With or without filtering: alternatives. HP Filter / Polynomial detrending.
(b)Expectations-based approach Identification ex-post of uncertainty ex-ante of economic agents. Expectationbased, detectingbreaks and removingregularities: Detrendedresiduals + decycling b.1) Detrending + decycling b.2) Forecasting errors (the best you can do)
First expectations adjustment: detrending “Much care has to be dedicated to the detrending procedure since a wrong specification can bias severely the subsequent analysis” (Bee Dagum) “Different detrending procedures are alternative windows which look at the series from different perspectives” (Canova) • Bee Dagumet al. (2006), “A critical investigation on detrending procedures for non-linear processes”, J. of Macroeconomics (vol 28). • Kauermannet al. (2011), "Filtering time series withpenalizedsplines", Studies in Nonlinear Dynamics and Econometrics, (vol 15(2)) • Canova (1998), “Detrending and business cycle facts: A user’s guide”, Journal of Monetary Economics (vol 41).
b.1) Detrending + decycling Distinction between variability and volatility. Implicit assumptions about decomposition of data: knowledge and ignorance: agents perceive regular but not irregular movements of economic time series. Unexpected portion, the unpredictable component of variability. • SD of Hodrick Prescott (HP) filtered residuals • SD of polynomial detrending residuals
Decycling: Fourier decomposition Bolch and Huang Periodic components of a time series
Choice of thebestmethod Thebestempiricalmethodshouldbedeterminedbythemodeling of economicagents´choices and thechannels of effectsonactivity and distribution. Butthereisnot a canonical modeltotake as a reference. Forempiricalmeasuring TOT volatility: Volatilityisassociatedwithuncertainty. TOT fluctuations are exogenous in thesmall open economy (SOE)
(b2) Expectations-based approach Previous methods suffer a temporal inconsistency Is tackled through: b.2) Forecasting errors (the best you can do) Out of sampleestimation and errors
V Empirical identification of GDP and TOT volatility. Temporal volatility profiles for Argentina: stylized facts. Cathegories to compare: amplitude, breaks, asymmetry, thresholds …
Modeling and estimatinguncertainty (3) Original Series Detrended Residuals Detrended + Decycled Residuals Volatility HP Filter / Polynomial Detrending Fourier Decomposition Standard Deviation
Data and methods Data: Argentina TOT and GDP loggedfromindex 1993=100. 1810 – 2012 (Ferreres & INDEC) Throughouttheexercisesall series are logged Detrending Cubicpolynomialdetrending HP filterdetrending (lambda = 100) Detrending + decycling Fourier decomposition Time inconsistency and thebestyou can do. Out of sampleforecasting
Statisticalmeasuresfor Argentina: single SD and 30 previousyearsrollingwindowRW • TOT, fourcomparativegraphs • Figure V.1. Statisticalapproach. A single SD of logged TOT and GDP forthewholeperiod. • Figure V.2. Statisticalapproach and expectationsapproach: detrendedwithbreaks. SD of 30 previousyears RW representsobserved data and perception of the data generatingprocess DGP. • Figure V.3. Detrendedwithbreaks + decycling • Figure V.4. Thebestyou can do • GDP onlydetrending • Figure V.3. Expectationsapproach