1. Chapter 4 Sources of Macroeconomic Fluctuations
© Pierre-Richard Agénor and Peter J. Montiel

2. Macroeconomic shocks and their propagation mechanisms are likely to differ in developing countries.
This chapter based on Agénor, McDermott and Prasad (1997).
Analysis of business cycle regularities is based on quarterly data for a group of twelve middle-income countries.
These are Colombia, Chile, India, Korea, Malaysia, Mexico, Morocco, Nigeria, the Philippines, Tunisia, Turkey, and Uruguay.

3. The Data.
Detrending Techniques.
Assessing Macroeconomic Fluctuations.
Summary of the Findings.

4. The Data

5. Figure 4.1: information on a number of key economic characteristics of the sample of countries.
Most of these countries could be characterized as middle-income countries.
Urbanization rates and the proportions of agricultural output as a share of total GDP indicate that agriculture is an important, but not dominant, sector.
Indices of industrial output are used to measure business cycle fluctuations.
Manufacturing sector accounts for a significant fraction of total GDP.

12. This variable is a reasonable proxy for measuring the aggregate cycle, because it
corresponds to output in the traded goods sector,
is closely related to business cycle shocks, either exogenous or policy-determined.
None of these countries had sustained episodes of hyperinflation during this period.
Export growth is an important contributor to overall GDP growth.

13. Detrending Techniques

14. Economic fluctuations at business cycle frequencies are examined by decomposing all the macroeconomic series into
nonstationary (trend);
stationary (cyclical) components.
Reason: certain empirical characterizations of the data are valid only if the data are stationary.
Stationary components obtained using different filters display different time series properties.
In order to examine the robustness of their results, three alternative filters have been used.

15. The Hodrick-Prescott Filter.
The Band-Pass Filter.
The Nonparametric Method.

16. The Hodrick-Prescott Filter
Seasonally adjusted variable xt can be written as the sum of an unobserved trend component, xt*, and a residual cyclical component, xtc:
xt = xt* + xtc.
Standard HP filter: trend component moves continuously and adjusts gradually.
xt* is extracted by solving the following minimization problem:
min
xt* 
t=1 T
(xt–xt*)2 + 
t=2
T-1
[(xt+1*–xt*) – (xt*–xt-1*)]2

17. Lagrange multiplier  is a positive number that penalizes changes in the trend component.
The larger is , the smoother is the resulting trend series.
Criticisms:
It removes valuable information from time series, and imparts spurious cyclical patterns to the data.
Choice of the value of : set to 1600 for quarterly time series.
This may reflect an overly stringent implicit assumption about the degree of persistence in xt.
Agénor, McDermott and Prasad (1997) choose a value of  for each individual series generalized cross-validation.

18. How is it applied?
Leave the data points out one at a time.
Choose the value of the smoothing parameter under which the missing data points are best predicted by the remainder of the data.
Estimates of the smoothing parameter showed a wide range of variation across countries and across data series.

19. The Band-Pass Filter Developed by Baxter and King (1995).
It is a moving average that filters both high frequency “noise” and low frequency “trends”.
It is constructed by
combining a low-pass filter and a high-pass filter;
imposing constraints that eliminate fluctuations at frequencies higher and lower than those corresponding to typical business cycle frequencies.
Frequency cut-offs correspond to 6 quarters and 32 quarters.

20. The Nonparametric Method
No specification of functional form of the trend component of the underlying series or degree of smoothing applied to the actual data.
It permits the modeling of trends that involve higher-order polynomials without imposing a particular functional form on the trend component.

21. Assessing Macroeconomic Fluctuations

22. Measure the degree of comovement of a series yt with industrial output xt by the magnitude of the correlation coefficient (j), j  {0, 1, 2, …}.
These correlations are between the stationary components of yt and xt, with both components derived using the same filter.
yt is procyclical, acyclical, or countercyclical, depending on whether the contemporaneous correlation coefficient (0) is positive, zero, or negative.
yt is strongly contemporaneously correlated if  |(0)| < 1.
yt is weakly contemporaneously correlated if  |(0)| < 0.26.
yt is contemporaneously uncorrelated with the cycle if  |(0)| < 0.13.

23. (j) indicates the phase-shift of yt relative to the cycle in industrial output.
yt leads the cycle by j period(s) if |(j)| is maximum for a positive j, is synchronous if |(j)| is maximum for j = 0, and lags the cycle if |(j)| is maximum for a negative j.
Cross-correlations between domestic output and the following variables:
variables that represent economic activity in the main industrial countries and the world real interest rate;
public sector expenditure and revenues;
real wages;
prices and inflation;
monetary aggregates, monetary velocity, and domestic credit to the private sector;

24. foreign trade variables (fluctuations in merchandise trade and terms of trade);
nominal and real effective exchange rates.

25. Summary of the Findings

26. Cross-correlations over the period 1978:Q1 to 1995:Q4.
Main findings can be summarized as follows:
Output volatility (standard deviations of the filtered cyclical component of industrial production), varies substantially across developing countries.
On average, it is higher than the level observed in industrial countries.
There is also considerable persistence in output fluctuations in developing countries.
Activity in industrial countries has a positive but relatively weak influence on output in developing countries.
Real interest rates in industrial countries are positively associated with output fluctuations.

27. Government expenditure is countercyclical.
Government revenues are acyclical in some countries, and significantly countercyclical in others.
Fiscal impulse is negatively correlated with the business cycle.
Cyclical behavior of nominal wages varies across countries and is not robust across filters.
Evidence strongly supports the assumption of procyclical real wages.
There is no consistent relationship between the stationary components of the levels of output and prices, or the levels of output and inflation.
Contemporaneous correlations between money and output are positive, but not strong.

28. Contemporaneous correlations between the velocity of broad money and industrial output are strongly negative across all filters for almost all the countries.
Domestic credit and industrial output are positively associated for some countries.  
No robust correlation between merchandise trade movements and output.
Cyclical movements in the terms of trade are strongly and positively correlated with output fluctuations.
No systematic patterns in the contemporaneous correlations between nominal effective exchange rates and industrial output.
Overall result: Importance of supply-side shocks in driving business cycles in developing countries.

29. Problems:
Using cross-correlation coefficients as indicators for evaluating the empirical relevance of demand-oriented, versus supply-oriented, macroeconomic theories can be problematic.
Results are not uniform across countries.