How Important Are Terms of Trade Shocks?

Size: px

Start display at page:

Download "How Important Are Terms of Trade Shocks?"

Jeffery Elliott
6 years ago
Views:

1 How Important Are Terms of Trade Shocks? Stephanie Schmitt-Grohé Martín Uribe Columbia University October 28, 215 1

2 Conventional View: Terms-of-trade shocks are a major source of business-cycle fluctuations in poor and emerging countries. (ex: Mendoza, IER 1995; Kose, 22.) This paper: Terms-of-trade shocks explain on average only about 1 percent of the variance of output in poor and emerging countries. 2

3 The Terms of Trade: tot t = P x t P m t tot t = terms of trade. Pt x = price of exports. Pt m = price of imports. 3

4 Mendoza (IER, 1995): Results show that terms-of-trade shocks account for nearly 1/2 of actual GDP variability. Methodology: 1. Collect annual terms-of-trade data from 23 developing countries. Sample period is 196 to 199. HP filter log of tot t. Compute serial correlation and standard deviation of tot t. Then take averages (means) across countries. This yields: corr(tot t, tot t 1 ) =.414 std(tot t ) = Build a theoretical model with 3 sectors, exportables, importables, and nontradables. Calibrate the model. Assume that tot t is exogenous to the country and follows a univariate process with serial correlation.414 and standard deviation

5 3. Compute the standard deviation of output under the assumption that terms-of-trade shocks are the only source of uncertainty. This yields std(gdp model t ). 4. Compare std(gdpt model ) with the cross-country average standard deviation of GDP observed in his panel. Mendoza (1995) finds std(gdp model t ) std(gdp data t ) =.56 Thus, terms-of-trade shocks explain 31 percent (= ) of the variance of GDP. 5

6 Observations on Standard Methodology The finding that TOT shocks explain over 1/3 of the variance of output is obtained in the context of highly stylized structural model (essentially, an open-economy version of the RBC model). The methodology relies on the theoretical model representing a reasonable approximation to the actual transmission mechanism of TOT shocks. This, aspect, however, is typically unexplored. An Alternative Methodology: SVAR Analysis The SVAR approach presented here uses the same restrictions to identify TOT shocks, but allows for a more flexible specification of the transmission mechanism. 6

7 Part I: SVAR Analysis 7

8 Empirical Model Let x t = [ tot t, tb t, ŷ t, ĉ t, ît, RER t ]. VAR: x t = A x t 1 + u t Identification of ToT shocks: u t = Πɛ t ɛ t (, I) Π 1,j = for j = 2,...,6 ToT process is univariate: A 1,j = for j = 2,...,6 All variables (except tb t ) are log-quadratically detrended. Estimate the SVAR country-by-country using OLS. 8

9 Data: Include all poor and emerging countries that have at least 3 consecutive annual observations of output, consumption, investment, net exports, the terms of trade, and the real exchange rate in the World Bank s WDI database. Poor and emerging countries are defined as countries with average PPP-converted GDP per capita in U.S. dollars of 25 over the period 199 to 29 below 25, dollars. 38 countries satisfy both criteria. Algeria, Argentina, Bolivia, Botswana, Brazil, Burundi, Cameroon, Central African Republic, Colombia, Congo, Dem. Rep., Costa Rica, Cote d Ivoire, Dominican Republic, Egypt, Arab Rep., El Salvador, Ghana, Guatemala, Honduras, India, Indonesia, Jordan, Kenya, Korea, Rep., Madagascar, Malaysia, Mauritius, Mexico, Morocco, Pakistan, Paraguay, Peru, Philippines, Senegal, South Africa, Sudan, Thailand, Turkey, and Uruguay. 9

10 Sample period: (32 years). Our sample of 38 countries. tot t = ρ tot t 1 + σ tot ɛ tot t ; ɛ tot t (,1) Estimate ρ and σ tot country by country σ ρ tot 1 ρ 2 Median.52.1 Interquartile Range [.41,.61] [.9,.13] 1

11 Impulse Response to A 1% Increase in the Terms of Trade SVAR Evidence, Median across 38 countries 1 Terms of Trade.6 Trade Balance.5 Output % dev. from trend % dev. from GDP trend % dev. from trend Consumption 1.2 Investment.5 Real Exchange Rate % dev. from trend % dev. from trend % dev. from trend

12 Observations on the estimation results: half-life of TOT shock is just 1 year. R 2 of tot equations is modest on average, 3 percent HLM effect (i.e., tb in response to a ToT appreciation) hold in 29 out of the 38 countries. On average, increase in GDP is.4 percent on impact. On average, c and i increase with a one-year delay. Tot appreciation leads to an appreciation of the real exchange rate. 12

13 Share of Variance Explained by Terms of Trade Shocks: SVAR Evidence tot tb y c i RER Median Median Absolute Deviation

14 Share of Variance of Output Explained by Terms of Trade Shocks Number of Countries 2 Algeria 18 Bolivia Burundi 16 C. African R. Congo, DR 14 Costa Rica El Salvador 12 Ghana Brazil Guatemala Cameroon 1 Honduras Colombia Kenya Dominican R. 8 Korea, Rep. India Madagascar Jordan 6 Malaysia Mexico Mauritius Peru 4 Morocco Philippines Pakistan South Africa 2 Paraguay Thailand Argentina Sudan Botswana Senegal Turkey Indonesia Uruguay Cote d Ivoire Egypt 1% 2% 3% 4% 5% 6% Share of Variance of Output Explained by TOT Shocks (percent) 14

15 Summary of SVAR Approach On average, TOT shocks explain 1 percent of the variance of output in poor and emerging countries. In only 5 countries (Botswana, Egypt, Cote d Ivoire, Sudan, and Uruguay) do ToT shocks explain more than 3 percent of the variance of output. Thus, the SVAR evidence is at odds with the conventional wisdom according to which ToT shocks account for a large share of output variability in poor and emerging markets. 15

16 Part II: Use a structural model to assess the importance of ToT shocks. Estimate some model parameters country-by-country using data from same 38 emerging and poor countries used in the SVAR. 16

17 The Theoretical Model small open economy that takes terms of trade as given. 3 sectors of production: exportable goods, importable goods, nontradable goods, with variable capital and labor inputs. Similar to Mendoza (IER,1995) but with more flexibility: Capital in the nontraded sector can vary over time. Labor in the importable and exportable sectors can vary. Investment goods have a domestically produced component. Allow for sector specificity in capital and labor. 17

18 The household problem: Preferences: E t= β t U(c t, h m t, hx t, hn t ) Budget constraint: c t +i m t +ix t +in t +pτ t d t+φ m (k m t+1 km t )+Φ x(k x t+1 kx t )+Φ n(k n t+1 k n t ) = pτ t d t+1 1+r t + w m t hm t + w x t hx t + wn t hn t + um t km t + u x t kx t + un t kn t 18

19 Final Goods Production y f t = [ χ τ (a τ t) 1 1 µτn + (1 χ τ )(a n t ) 1 1 µτn ] µτn, yt f a τ t a n t = final goods. = composite of traded goods. = nontraded goods. µ τn = elasticity of substitution between T and N goods. χ τ = expenditure share on tradables if µ τn. 19

20 Production of the Tradable Composite Good a τ t = [ χ m (a m t ) 1 1 µmx + (1 χ m )(a x t) 1 1 µmx ] µmx a τ t = composite of traded goods. a m t = importable goods. a x t = exportable goods. µ mx = elasticity of substitution between importables and exportables. χ m = expenditure share if µ mx = 1. 2

21 Production of Importable Goods y m t = A m (k m ) αm (h m ) 1 α m y m t = quantity of importable goods produced domestically. A m = level of productivity in the importable sector. k m t = capital input in the importable sector. h m t = labor input in the importable sector. 1 α m = labor share in the importable sector. 21

22 Production of Exportable Goods y x t = Ax (k x ) αx (h x ) 1 α x yt x = quantity of exportable goods produced. A x = level of productivity in the exportable sector. kt x = capital input in the exportable sector. h x t = labor input in the exportable sector. 1 α x = labor share in the exportable sector. 22

23 Production of Nontradable Goods y n t = A n (k n ) αn (h n ) 1 α n yt n = quantity of nontraded goods produced. A n = level of productivity in the nontradable sector. kt n = capital input in the nontradable sector. h n t = labor input in the nontradable sector. 1 α n = labor share in the nontraded sector. 23

24 To ensure a stationary equilibrium process for external debt, we assume that the country interest-rate premium is debt elastic, r t = r + p(d t+1 ) 24

25 Terms of Trade Process ( tott ln tot ) = ρln ( ) tott 1 tot + σ tot ɛ tot t ; ɛ tot t (,1) tot >. ρ ( 1,1). σ tot >. 25

26 Functional Form Assumptions p(d) = ψ ( e d d 1 ) Φ j (x) = φ j 2 x2 ; j = m,x, n 26

27 Calibrated and Estimated Parameters Calibrated Structural Parameters ρ σ tot α m, α x α n ω m, ω x, ω n µ mx µ τn tot A m, A n β σ δ r /(1 + r ) Moment Restrictions σ i σ tb σ im+ix p σ y σ y σ in s n s x s m y m tb p x y x Implied Structural Parameter Values φ m φ x φ n ψ χ m χ τ d A x β Notes. Country-specific estimates. σ i σ y and σ tb σ y s n p n y n /y, s x x/y, are conditional on tot shocks s tb (x m)/y, where y p m y m + p x y x + p n y n. 27

28 Key parameters determining the importance of terms of trade shocks: ρ and σ tot, the more volatile and the more persistent are terms of trade shocks, the more volatile is output. The size of the nontraded sector: pn y n y (= 5%). The larger the nontraded sector, the smaller the output effects of tot shocks. x+m The steady-state trade share: y (= 39%). The larger the trade share, the larger the output effects of tot shocks. 28

29 Estimate capital adjustment cost parameters and the debt elasticity of the interest rate, φ m, φ x, φ n, ψ, χ m, to match country-bycountry the relative standard deviations σ i /σ y and σ tb /σ y conditional on terms of trade shocks and σ ix +i m /σ in =

30 Medians of Country-Specific Estimates of the Capital Adjustment Cost Parameters and the Debt Elasticity of the Interest Rate σ i /σ y σ tb /σ y φ m φ x φ n ψ Data Model Data Model Median MAD

31 Median of Country-Specific Predicted Impulse Response to a Ten-Percent Terms-of-Trade Shock Terms of Trade 1 RER 2 p N Trade Balance Imports Exports 2 GDP 1 Consumption 4 Investment 2 y M y X 2 y N 2 i M 1 i X 6 i N

32 Observations of the figure: Substitution effect of an increase in tot t on supply side. Firms produce more exportables and less importables, and given p n t would produce less nontradables. Substitution effect of an increase in tot t on demand side. Demand for importable goods and nontraded goods rises, domestic demand for exportable goods falls. Wealth effect is positive increasing the demand for all goods. Price of nontradables, p n t, rises and the real exchange rate appreciates. Both exports and imports increase. Net effect on trade balance turns out to be positive. Thus model impulse response is consistent with Harberger-Laursen-Metzler effect. Aggregate investment increases by less than 1% on impact. But investment in the exportable sector rises by 61% while in decreases by over 4% in the importable sector. 32

33 Variance Decomposition: We construct the counterpart to the observable variable GDP at constant LCU in the theoretical model as: y constant prices t = p x ss yx t + pm ss ym t + p n ss yn t, where p i ss denotes the steady-state price of good i = x, m, n. In the theoretical model, we deflate consumption and investment by the GDP deflator (constructed as a Paasche index): P t = px t yx t + pm t ym t + p n t yn t p x ss yx t + pm ss ym t + p n ss yn t 33

34 Finding: The median share of the variance of output explained by tot shocks, according to the 38 calibrated models is 13 percent. This number is close to the SVAR results but far from the conventional wisdom. 34

35 Median Share of Variance Explained by Terms of Trade Shocks tb y c i RER Theoretical Model SVAR Model

36 Cross Country Variation in the Share of Variance Explained by Terms of Trade Shocks in the Theoretical Model tot tb y c i RER Cross-country Median Median Absolute Deviation

37 Evaluating The Conventional Wisdom How Does the Model Fit Country-Level Data? 6 Cen Sen Share of Output Variance Due to ToT Shocks Ind Mex 5 Model 4 Hon 3 Gua Mad El 2 Ken Bol Alg Pak Gha Par Ind Egy Sud Cot Col Bra Tur 1 Bur Cos Mor Phi Mau Dom Uru Sou Arg Per Bot Con Mal Kor Jor Cam Tha Data The RBC framework appears to have difficulty capturing the propagation of TOT shocks. 37

38 Variance of Consumption, Investment, the Trade Balance, and the Real Exchange Rate Explained By Terms-of-Trade Shocks: SVAR Versus Model Consumption Investment Theoretical Model Theoretical Model SVAR Model SVAR Model 6 Real Exchange Rate 6 Trade Balance 5 5 Theoretical Model Theoretical Model SVAR Model SVAR Model 38

39 Conclusion 1. Conventional wisdom has it that terms of trade shocks represent a major source of fluctuations for emerging countries. 2. Using SVAR analysis, the present study finds a modest role for TOT shocks. 3. The analysis suggests that the theoretical framework on which the conventional wisdom is based, an open economy version of the RBC model, fails to capture well the transmission mechanism of TOT shocks at the country level. 39

40 Extras 4

41 Model Predictions Using the Median ToT Process and Matching Median Moments φ m φ x φ n ψ ρ σ tot Share of Variance Explained by Terms of Trade Shocks tb y c i RER SVAR Model Theoretical Model

42 tot t = a 11 tot t 1 + π 11 ɛ 1 t ; ɛ1 t (,1) Table 1: The Terms of Trade Process: Country-by- Country Estimates Country a 11 π 11 R 2 Algeria Argentina Bolivia Botswana Brazil Burundi Cameroon Central African Republic Colombia Congo, Dem. Rep Costa Rica (continued on next page) 42

43 Table 1 (continued from previous page) Country a 11 π 11 R 2 Cote d Ivoire Dominican Republic Egypt, Arab Rep El Salvador Ghana Guatemala Honduras India Indonesia Jordan Kenya Korea, Rep Madagascar Malaysia Mauritius Mexico Morocco (continued on next page)

44 Table 1 (continued from previous page) Country a 11 π 11 R 2 Pakistan Paraguay Peru Philippines Senegal South Africa Sudan Thailand Turkey Uruguay Median MAD

45 Table 2: Share of Variance Explained by Terms of Trade Shocks: Country- Level SVAR Evidence Country tot tb y c i RER Algeria Argentina Bolivia Botswana Brazil Burundi Cameroon Central African Republic Colombia Congo, Dem. Rep Costa Rica Cote d Ivoire (continued on next page) 43

46 Table 2 (continued from previous page) Country tot tb y c i RER Dominican Republic Egypt, Arab Rep El Salvador Ghana Guatemala Honduras India Indonesia Jordan Kenya Korea, Rep Madagascar Malaysia Mauritius Mexico Morocco Pakistan (continued on next page)

47 Table 2 (continued from previous page) Country tot tb y c i RER Paraguay Peru Philippines Senegal South Africa Sudan Thailand Turkey Uruguay Median Median Absolute Deviation Note. Shares are expressed in percent.

48 Table 3: Country-Specific Estimates of the Capital Adjustment Cost Parameters and the Debt Elasticity of the Interest Rate σ i /σ y σ tb /σ y Country φ m φ x φ n ψ Data Model Data Model Algeria Argentina Bolivia Botswana Brazil Burundi Cameroon Central African Republic Colombia Congo, Dem. Rep Costa Rica Cote d Ivoire Dominican Republic Egypt, Arab Rep El Salvador Ghana Guatemala Honduras India Indonesia Jordan Kenya Korea, Rep Madagascar Malaysia (continued on next page) 44

49 Table 3 (continued from previous page) σ i /σ y σ tb /σ y Country φ m φ x φ n ψ Data Model Data Model Mauritius Mexico Morocco Pakistan Paraguay Peru Philippines Senegal South Africa Sudan Thailand Turkey Uruguay Median Median Absolute Deviation

50 Table 4: Share of Variance Explained by Terms of Trade Shocks: Country Level Predictions of the Theoretical and SVAR Models tb y c i rer Country TH SVAR TH SVAR TH SVAR TH SVAR TH SVAR Algeria Argentina Bolivia Botswana Brazil Burundi Cameroon CAR Colombia Congo Costa Rica Cote d Ivoire Dom Rep Egypt El Salvador Ghana Guatemala Honduras India Indonesia Jordan Kenya Korea, Rep Madagascar Malaysia (continued on next page) 45

51 Table 4 (continued from previous page) tb y c i rer Country TH SVAR TH SVAR TH SVAR TH SVAR TH SVAR Mauritius Mexico Morocco Pakistan Paraguay Peru Philippines Senegal South Africa Sudan Thailand Turkey Uruguay Median MAD

slides chapter 8 Nontradable Goods And The Real Exchange Rate

slides chapter 8 Nontradable Goods And The Real Exchange Rate Princeton University Press, 217 Introduction The MX model of Chapter 7 overestimates the role of terms-oftrade shocks as drivers of the business