Structural breaks, unit root tests and long time series
Download as PPT, PDF2 likes2,966 views
The document discusses the analysis of macroeconomic variables such as inflation, unemployment, and interest rates, emphasizing their interconnectivity and impact on economic stability. It highlights historical cases, such as hyperinflation in Zimbabwe and the Great Recession in the U.S., to illustrate how high inflation and economic shocks can lead to long-term adverse effects. Additionally, it addresses the importance of time series analysis in understanding these relationships and the potential pitfalls of misleading unit root tests in the presence of structural breaks.
Structural breaks, unit root tests and long time series
- 2. Macro economic variables consist of GNP, unemployment, inflation, interest rate, exchange rate, balance of payments, etc. Unacceptable levels (think high inflation) or instability (think alternating periods of high and low growth) in any of the above variables can be very distressing for the people (think prices increasing by 20% every month OR being fired from your job due to low revenue growth prospects of your employer due to recession)
- 3. Hyperinflation in Zimbabwe: ! From NewYorkTimes, “at a supermarket near the centre of this tatterdemalion capital, toilet paper costs $417. No, not per roll. Four hundred seventeen Zimbabwean dollars is the value of a single two-ply sheet. A roll costs $145,750 — in American currency, about 69 cents.The price of toilet paper, like everything else here, soars almost daily, spawning jokes about an impending better use for Zimbabwe's $500 bill, now the smallest in circulation.” ! 100Trillion Zimbabwean dollar bills were in circulation ! In November 2008, inflation hit a high of 79.6 billion percent. So $Z 79600000000 would have to be paid in Nov 2008 for a pen priced at $Z1 in Nov 2007
- 4. The Great Recession (2007-2009) 1. US unemployment rate rose from 5% in 2008 pre-crisis to 10% by late 2009 2. Post recession Income levels of the median male worker was down to 1968 levels 3. Approximately 5.4 million people have been added to federal disability rolls 4. U.S. total national debt rose from 66% GDP in 2008 pre-crisis to over 103% by the end of 2012
- 5. Macro variables reflect different aspects of the same economy so they are interconnected and fluctuations in one can quickly translate into fluctuations in the others Consider this situation: When inflation is high, people may lose confidence in money as the real value of savings is severely reduced This discourages savings due to the fact that the money is worth more presently than in the future This expectation reduces economic growth because the economy needs a certain level of savings to finance investments which boosts economic growth. Also, inflation makes it harder for businesses to plan for the future. It is very difficult to decide how much to produce, because businesses cannot predict the demand for their product at the higher prices they will have to charge in order to cover their costs. Savings Growth Investment
- 6. To stabilize the economy over time, governments need to formulate policy to control the macro variables for which they need to understand the long term relation between them. Is this possible? The most powerful tool to understand the relation between crucial macro variables is Time Series Analysis: a branch of Econometrics or statistical analysis of economic variables
- 7. • Nelson and Plosser (1982) argued that almost all macroeconomic time series, have a unit root What does this mean: • In the absence of unit root (stationary), the series fluctuates around a constant long- run mean and implies that the series has a finite variance which does not depend on time. • On the other hand, non-stationary series have no tendency to return to long-run deterministic path and the variance of the series is time dependent. • Non-stationary series suffer permanent effects from random shocks and thus the series follow a random walk • Think tourist arrivals at a destination over time. If this series is non-stationary, then in case of random shocks like terrorist attacks or natural disaster, the number of tourist arrivals never revert to their original mean, but if the series was stationary they would have.
- 8. • If this were true, there is no use to policy anymore. The effects of hyperinflation or a recession on the economy are permanent and incurable and we are doomed to be on a low level path forever. • Such a Greek tragedy scenario where you are completely at the mercy of the Gods doesn’t seem relevant for current times
- 9. Perron (1989), argued that in the presence of a structural break, the standard ADF tests for unit root are biased towards the non-rejection of the null hypothesis The series on the left is non-stationary but the one on the right is not.. However, ADF tests might misleadingly point out the series on the right to be non-stationary as well
- 10. Testing for structural breaks is extremely important while analyzing long time series. Otherwise, all subsequent analysis might be misleading For instance, two series are cointegrated if they are individually I(1), but some vector of coefficients exists to form a stationary linear combination of them However, in the presence of structural breaks, unless proper testing is done, the individual series might mistakenly by labeled I(1)
- 11. Test Model Software Perron (1989)** Exogenous with one break Zivot and Andrews (1992)* Endogenous with one break Eviews Lumsdaine and Papell (1997)* Endogenous with two breaks GAUSS Lee and Strazicich (2003)** Endogenous with two breaks RATS Gregory and Hansen (1996) One Endogenous break in cointegration framework Eviews Saikkonen and Lütkepohl (2000) One Endogenous break in cointegration framework GAUSS Bai and Perron (2003) Endogenous multiple breaks RATS,Eviews * Assume no break(s) under the null hypothesis of unit root ** Assume break(s) under both the null and the alternative hypothesis
- 12. The Indian Ministry of Statistics and Program Implementation has just introduced a structural break in the GDP series. Read up and be careful The tests in the earlier slide are quite technical. But expectedly anyone interested in this issue is likely to have a technical appetite. Hence happy reading. Zivot & Andrews (intensely mathematical) is good place to start Tests for panel data are a different set. Search Westlund, Levin-Lin-Chu, Pedroni etc.