UNIVERSITY OF ILLINOIS LIBRARY AT URBANA CHAMPAIGN BOOKSTACKS Digitized by the Internet Archive in 2012 with funding from University of Illinois Urbana-Champaign http://www.archive.org/details/persistenceofinn91128crib Faculty Working Paper 91-0128 Graduate Student Series #75 330 B385 1991:128 COPY 2 Persistence of Innovations and Economic Policy: The Brazilian Experience The Library of the MAY I o ]v>| University of Minds of Urbana-ChampaifK; Francisco Cribari-Neto Department of Economics Bureau of Economic and Business Research College of Commerce and Business Administration University of Illinois at Urbana-Champaign BEBR FACULTY WORKING PAPER NO. 91-0128 College of Commerce and Business Administration University of Illinois at Urbana-Champaign April 1991 Persistence of Innovations and Economic Policy. The Brazilian Experience Francisco Cribari-Netof Department of Economics I am especially thankful to William Maloney for his careful readings of several earlier drafts and insightful suggestions. Also, thanks go to Ana Dolores Novaes and Carlos Brandao Cavalcanti for their helpful comments. I benefitted from conversations with Paul Newbold as well. The usual disclaimers apply. f From the Ph.D. Program in Economics at the University of Illinois at Urbana-Champaign. Persistence of Innovations and Economic Policy: The Brazilian Experience Francisco Cribari-Neto April 1991 - Abstract - According to the conventional view, shocks in output have effects on the short-run, but are not persistent in the long-run. However, it has been argued in the recent literature that shocks in output do have long-run effects. This paper utilizes a nonparametric technique in order to check whether innovations in Brazilian GDP are persistent. We find that an innovation of one percent today is more than proportionately incorporated in the long-run, and we call this situation shock dominance. We also examine whether economic policies can alter the persistence of shocks. Our results indicate that on some occasions stabilization policies do have long-run effects. t I am especially thankful to William Maloney for his careful readings of several earlier drafts and insightful suggestions. Also, thanks go to Ana Dolores Novaes and Carlos Brandao Cavalcanti for their helpful comments. I benefited from conversations with Paul Newbold as well. The usual disclaimers apply. $ From the PhD Program in Economics at the University of Illinois at Urbana-Champaign. I. Introduction: One of the most controversial questions in Macroeconomics is the one related to the persistence of innovations in long-run GDP (Gross Domestic Product). The more conservative view is the one related to deterministic trend-cycles decomposition. According to this view, product movements can be separated into two different components, trend and cycles, where the first one follows a deterministic path and the second obeys some stationary stochastic process. This view has been challenged since the publication of the Beveridge and Nelson (1981) and Nelson and Plosser's (1982) influential papers. According to their understanding, the output trend itself follows a stochastic process. Although both cases do not seem very different, the implications of the stochastic trends are far-reaching. Among these implications, four can be highlighted. The first one is related to the permanence of innovations. Unlike the deterministic case where innovations are temporary, in the second scheme we can have a high degree of permanence of shocks in the current product. Campbell and Mankiw's results (1987) indicate that an innovation of one percent in the present output is incorporated by more than one percent in the long-run output level. The second important implication is related to the relationship between short and long-run policies. According to the traditional view, stabilization policies do not have long-run effects, while the second understanding implies that non-anticipated policies do have effects on the long-run behavior of the product. Roughly speaking, this means that economic agents are forced to live with the effects of unsuccessful stabilization policies in any period of t ime. Thirdly, there is the issue of measurement of cycles. The traditional view tends to overestimate the cyclical component once any stochastic movement is attributed to it, as shown by Cuddington and Urzua (1989) for the Colombian case. Finally, if the trend in output is stochastic and one detrends the data assuming that it is deterministic some spurious patterns are introduced. In particular, the autocorrelation function from detrended data may show pseudocycl ical movements, as shown by Nelson and Kang (1981). Moreover, one is also limiting uncertainty, reducing the original variance of data, and reducing the influence of past over the future. This point is particularly important since detrending variables before the analysis or 2 including time as an explanatory variable in regressions seems to be a common practice in applied works. In the present paper we shall try to answer the following question: To what extent are the innovations in the Brazilian GDP persistent? We shall argue that, based on our results, this permanence occurs to a high degree. Next, we try to establish a connection between persistence of innovations and economic policy. With this aim we analyze the Brazilian experience after the first international shock in oil prices. The rest of the paper is organized in the following way. Section II briefly discuss some of the possible approaches to this issue. Section III sums up the results and discuss their significance to economic fields. Section IV covers the relation between economic policy and persistence of innovations. Finally, the last section summarizes some important implications resulting from our results. II. Methodology : The issue of persistence of innovations can be addressed using 3 two methodologies. The first one is the parametric approach, where impulse response functions are calculated from ARIMA models See also Chan, Hayya and Ord (1977), and Nelson and Kang (1984). 2 In fact these two practices are equivalent. 3 There are however several approaches. For instance, see Harvey (1984), Watson (1986) and Clark (1987) for structural models. For an analysis based on Markovian Processes see Hamilton (1989). previously estimated [Campbell and Mankiw (1987)]. Let Y be the logarithm of GDP. Assuming that its first difference is stationary of second order, and modelling it as an ARIMA we have: 0(B)(1-B)Y = 6(B)€ (1) where B is the Backward Shift Operator and e is white noise. From (1) we get: Y = 0(B)e (2. where 0(B) = (1-B) ^(B)] 1 6(B). Now we have that 0= = E 0, ( 3 ^ j=0 J The limit of 0. is the sum of coefficients of an infinite moving average representation for the level of Y , denoted by A(l). Then, for large i_ 0. is a measure of persistence, since it is nothing more than the accumulation of the effects of an innovation in t. This parametric approach, however, has some limitations. One of them is related to model specification. Like any parametric approach, there is no consensus in this field at all about what is the best method of model selection. Another serious restriction is the fact that two different models that seem to describe data equally well can give widely differing results for the long-run s 5 4 persistence of innovations. Furthermore, this type of analysis often leads to an overest imat ion of the long-run persistence. 4 See e. g. , Watson ( 1986 5 See Cochrane ( 1988) . The second approach is to use a nonparametric measure, such as the one developed in Cochrane (1988). He considers two different processes, one with deterministic trend (eq. 4), and the other a random walk with drift /i (eq. 5): Y4. = bt + V X .€, 4) Y = u + Y + € t M t-1 t (5: where € is white noise. Taking the variance of Y in both cases, and dividing it by k times the variance of first difference we get: V k = (6: for the deterministic case, and: v k Var (€ t> (7) Var (AY ) for the random walk. It is then possible to see that: Lim = k=x» in the deterministic case in the random walk case In fact there is no consensus in the literature about what is the most appropriate measure of long-run persistence. See e.g., Stock and Watson ( 1988). Thus it is possible to use Cochrane' s nonparametric measure to quantify the persistence of an innovation in log GDP. It can take any value ranging from (the deterministic case) up to more than 7 one. That is: V R . Var ' Wk' (8) kVar (Y t -Y ) is, for large k, a measure of how persistent the innovations are. An estimator for V is given by: k-1 V, 1 + 2 £ k - j p (9 ~ 1 : J where p. is the j'th sample autocorrelation coefficient from the u first difference of the series. One clear advantage of this analysis is that one has not to choose among competing models. Furthermore, it does not depend on the correlation between innovations in the secular component and in g the cyclical component of GDP. However, although this nonparametric measure does not encounter problems like model specification, it has its limitations as well. There is an explicit relationship between Cochrane' s measure and Campbel 1-Mankiw' s 10 impulse response function. Moreover, one has to be especially 7 Of course it is 1 in the random walk case. 8 A bias correction of n/(n-k+l) is usually applied, since there is a downward bias for large values of k. g See Cochrane ( 1988, p. 904 ) . 10 T 2 1/2 2 It is easily shown that A(l) = [V/U-R )] . where R is the fraction of variance that can be predicted from past information, 2 2 2 2 i.e., R = 1 - (o ^ tj c« f* C* W »H « « >H «1 - Therefore, the Brazilian experience is a first indication that there is a relationship between economic policy and shock incorporation. We have tried to investigate the effect of a policy under very special conditions, that is, in the situation where the shock seems to affect mainly the investment and where the government reaches its target of reducing the uncertainty and minimizing the effects on total investment, without generating problems capable of offsetting the achievement of those goals. IV. Implications and Conclusions: We can highlight several of the important implications of the results presented here. First, one has to be careful in measuring cycles, since the traditional methods tend to overestimate this component. Secondly, innovations in the Brazilian GDP are highly permanent. An innovation is incorporated by more than its original magnitude. Finally, we investigated the influence of economic policy on the persistence of innovations using the Brazilian experience after the first international shock in oil prices. Our results show that the Government was capable of reducing the persistence of that shock in the GDP. At that time the economic policy reduced uncertainty and the effects of the shock on investment. This fact leads us to conclude that under special circumstances there is a relationship between economic policy and persistence of innovations. Therefore, on some occasions, stabilization policies do have long-run effects. 15 References: BAER, W. (1989) The Brazilian Economy. Praeger, 3rd ed. BEVERIDGE, S. and C.NELSON (1981) "A New Approach to Decomposition of Economic Time into Permanent and Transitory Components with Particular Attention to Measurement of the 'Business Cycle' . " Journal of Monetary Economics, 7, 151-74. CAMPBELL, J.Y. and N.G.MANKIW (1987) "Are Output Fluctuations Transitory?" The Quarterly Journal of Economics, 102, 857-80. CARDOSO, E. and R. DORNBUSCH (1988) "Divida Bras i lei ra: Requiem para a Politica de Muddling Through." Pesquisa e Planejamento Economico, 18, 211-38. CHAN, K. , J.C.HAYYA and K. ORD (1977) "A Note on Trend Removal Methods: the Case of Polynomial Regression Versus Variate Differencing. " Econometr ica, 45, 737-44. CLARK, P.K. (1987) "The Cyclical Component of U.S. Economic Activity. " The Quarterly Journal of Economics, 102, 797-814. COCHRANE, J.H. (1988) "How Big is the Random Walk in GNP?" Journal of Political Economy, 96, 893-920. CUDDINGTON, J.T. and CM. URZUA (1989) "Trends and Cycles in Colombia's Real GDP and Fiscal Deficit." Journal of Development Economics, 30, 325-43. DICKEY, D.A. and W. A. FULLER (1979) "Distribution of the Estimators for Autoregressive Time Series with a Unit Root." Journal of the American Statistical Association, 74, 427-31. DIEB0LD, F.X. and G. D. RUDEBUSCH (1989) "Long Memory and Persistence in Aggregate Output. " Journal of Monetary Economics, 24, 189-209. FLAVIN, M. (1981) "The Adjustment of Consumption to Changing Expectations about Future Income." Journal of Political Economy, 89, 974-1009. FULLER, W. A. (1976) Introduction to Statistical Time Series. Wiley. HADDAD, C. (1977) "Crescimento do Produto Real Brasileiro: 1990/1947." J_n Formacao Economica do Brasil, Flavio Rabelo Versiani and Jose Roberto Mendonca de Barros (ed. ), Saraiva. HALL, R. E. (1978) "Stochastic Implications of the Life Cycle- Permanent Income Hypothesis: Theory and Evidence. " Journal of Political Economy, 86, 971-87. 16 (1986) "The Role of Consumption in Economic Fluctuations." in The American Business Cycle, R.J.Gordon (ed. ), The University of Chicago Press. HAMILTON, J.D. (1989) "A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle. " Econometrica, 57, 357-84. HARVEY, A.C. (1984) "Trends and Cycles in Macroeconomic Time Series. " Discussion Paper A45, Statistics Department, London School of Economics. INSTITUTO BRASILEIRO DE GEOGRAFIA E ESTATlSTICA (1987) Estatisticas Historicas do Brasil - Series Estatisticas Retrospect ivas, vol.3. IBGE, Rio de Janeiro. MANKIW, N.G. and M.D.SHAPIRO (1985) "Trends, Random Walks, and Permanent Income Hypothesis. " Journal of Monetary Economics, 16, 165-74. MARQUES, M. S. B. (1989) "A Politica Economica apos o Primeiro Choque do Petroleo e seu Impacto sobre as Trajetorias da Inflagao e do Hiato do Produto. " Revista Brasileira de Economia, 43, 239-72. MILLS, T.C. (1990) Time Series Techniques for Economists. Cambridge University Press. NELSON, C.R. (1987) "A Reappraisal of Recent Tests of the Permanent Income Hypothesis." Journal of Political Economy, 95, 641-46. NELSON, C.R. and H. KANG (1981) "Spurious Periodicity in Inappropriately Detrended Time Series. " Econometrica, 49, 741-51. . (1984) "Pitfalls in the Use of Time as an Explanatory Variable in Regression. " Journal of Business and Economic Statistics, 2, 73-82. NELSON, C.R. and C.PLOSSER (1982) "Trends and Random Walks in Macroeconomic Time Series. " Journal of Monetary Economics, 10, 139-62. STOCK, J.H. and M.W.WATSON (1988) "Variable Trends in Economic Time Series. " Journal of Economic Perspectives, 2, 147-74. TOMBINI, A. A. and P. NEWBOLD (forthcoming) "The Time Series Behavior of Brazilian Real Gross Domestic Product, 1947-1987: An Analysis of Interventions." World Development . WATSON, M.W. (1986) "Univariate Detrending Methods with Stochastic Trends. " Journal of Monetary Economics, 18, 49-75. 17 HECKMAN BINDERY INC. JUN95 , T D , ■ N. MANCHESTER, ound-To-Plras* |ND | ANA 46962