Business Cycle Mapping in SAARC Economies

I undertook a comparative econometrics analysis of business cycles in SAARC economies.
This platform contains research methodology and findings of the research paper. Readers who wish to obtain the complete text may reach me via farshad.bonyadih@gmail.com


1. Methodology
Focus of this paper is to see if various econometric models arrive at the same conclusion, or not. Principal trend-cycle methods include Hodrick Prescott trend-cycle decomposition, Beveridege Nelson trend-cycle decomposition and unobserved component model. Conclusions are based on average of cyclical components obtained from all three models.

Additionally, SupF and Bai and Perron tests are applied to see if there has been a structural change in evolution of the series. Both set of tests are applied to linear regression of yt on yt-1+1 and yt on 1. Econometric methods are applied to univariate autoregressive representation of log of real GDP following James H. Stock’s proposition on testing stability in reduced form forecasting methods (The econometric Analysis of Business Cycle). These datasets are obtained from Federal Reserve Bank of St. Loius.

Furthermore, to gauge the scope and strength of regional connectivity, 5 year rolling window binary correlation coefficients have been obtained between India and every other countries.


2. Conclusions
While all methods related to trend-cycle decomposition unanimously fit a reasonably good model series’ real Gross Domestic Product, they vary in modelling magnitude of cycles, slightly.
A. Bangladesh’s real gross domestic product demonstrate very high volatility from 1959-1975, moderate volatility till 1990 and rather minimal volatility henceforth. There is a stark reduction in standard deviation of decade-wise cyclical fluctuations over the past 4 decades. It indicates 1972, 1967, 1961, 1973 and 1969 to be extreme recessionary dates, while, 1970, 1960, 1971, 1974 and 1963 to have the highest peak points. Longest expansions occurred during 1997-2001, while longest recessions happened during 1991-1996 and 2002-2006 and 1965-1968. Structural break tests identify at least one break point that dates 1975.

B. India’s real gross domestic product shows high volatility during 1960-1990 and milder volatilities henceforth. 1966, 1979, 1974, 1957 and 2002 were years of extreme lows, while, 1974, 1969, 1978, 1977 and 1970 were years of high booms. Longest expansions occurred in 1969-1971, 1988-1990, 1998-2000 and 2005-2007. Longest recessions occurred in 1965-1968, 1984-1987,1991-1994 and 2001-2004. Structural break tests identify at least one break point—dating at 1988. Although models favor 4 break points too. Standard deviation and variances of average cyclical component increased in the 60s but decreased afterwards.

C. Cyclical component of Maldives’ RGDP demonstrate very high level of volatility during 1970-1990 and 2000-2010. Average of all three cycles indicate extreme recessions during 2005, 1982, 1976, 2001 and 1983. It also indicates high peaks in 2008, 1974, 2004, 1981 and 1980. Longest periods of recessions occurred during 1993-1996 and 1987-1989 and 1975-1977. Longest periods of growths occurred during 1997-2000. Standard deviation and variance of decadal-observations of the average cyclical components of all three model decrease up until 2000 but dramatically increase during 2000-2009. Structural break tests identify a single break at 1976.

D. Nepal’s real gross domestic product demonstrates high and more frequent volatility up until approximately 1995 and slightly milder and longer booms and busts. Longest recession lasted three years (1967-1969) and longest expansion lasted four years (1994-1997). 1983, 1980, 1973, 1963 and 1968 are years of extreme recessions, while, 1964, 1982, 2001, 1966 and 1979 are years of high booms. Standard deviation and variance of observations has decreased by 100% (in former) and 500% (in latter) in the 2000s compared to the 60s. Structural break tests identify at least one break in the series at around 1990.

E. Pakistan’s economy experienced longer and sharper phases of booms and busts till the 70s, milder but longer phases till the 90s and longer and sharper booms and busts in the 2000s. Averaged cyclical component indicates 1968, 1951, 2003, 2002 and 1960 and 1972 to be years of extreme recession, while 1970, 1953, 2006-2007, 1996 and 1992 having highest booms. Longest expansionary periods occurred in 2005-2009 and many other three-year long ones. Longest recessionary periods occurred in 1959-1964, 1976-1981, and 1983-1987 and 2010-2015. Standard deviation and variance of cyclical observations remains slightly steady till the 70s, decreases till the 90s and increases afterwards. Structural break tests confirm existence of at least one break at 1959.

F. Cyclical component of Sri Lanka shows extreme volatilities till mid 70s, and after 2000. From 70s till late 90s, the volatilities decrease to some extent. Longest recession occurred in 2001-2005, while longest expansions occurred in 1978-1982 and 1997-2000. 1956, 1953, 1957, 1966 and 2009 are years of extreme lows, while 1955, 1961, 1970, 2000 and 1952 are years of high booms. Standard deviation and variance of observations from the cyclical component historically has decreased, nonetheless, has a U-shape and increased from 1990 onwards. Structural break tests confirm existence of at least one break at 1959.

G. Maldives-India business cycles demonstrate mostly negative 5 year rolling correlation; Bangladesh-India business cycles exhibit both equally negative and positive 5 year rolling correlation with strong positive correlations during mid 60s and 70s, late 90s and late 2000s. Pakistan-India business cycles has been showing frequent and strongly positive, with mellow short-lasting negative 5 year rolling correlations. Nepal-India business cycles show short periods of medium positive and more frequently negative 5 year rolling correlation. Sri Lanka-India business cycles show short periods of medium positive and negative rolling correlation up until 2000 and the decade 2000-2010 characterizes strong and long positive correlation.

H. Cyclical components of these six countries do not have pair-wise correlation during 1970-2009. India and some countries exhibit correlation among decade-wise observations that are statistically significant at 5% level of confidence, during limited time horizons. These pairs are: In the 80s: India and Sri Lanka (-0.6); In the 90s: India and Maldives (-0.61); In 2000s: India and Pakistan (0.68).

0 comments:

Post a Comment

My Twitter Feed