BSc Hons Finance II BSc Hons Finance with

BSc (Hons) Finance II/ BSc (Hons) Finance with Law II Module: Principles of Financial Econometrics I Lecturer: Dr Baboo M Nowbutsing Topic 10: Autocorrelation Topic Nine Serial Correlation

Outline 1. 2. 3. 4. 5. 6. Topic Nine Introduction Causes of Autocorrelation OLS Estimation BLUE Estimator Consequences of using OLS Detecting Autocorrelation Serial Correlation

1. Introduction Ø Autocorrelation occurs in time-series studies when the errors associated with a given time period carry over into future time periods. Ø For example, if we are predicting the growth of stock dividends, an overestimate in one year is likely to lead to overestimates in succeeding years. Topic Nine Serial Correlation

1. Introduction Ø Times series data follow a natural ordering over time. Ø It is likely that such data exhibit intercorrelation, especially if the time interval between successive observations is short, such as weeks or days. Topic Nine Serial Correlation

1. Introduction Ø We expect stock market prices to move or move down for several days in succession. Ø In situation like this, the assumption of no auto or serial correlation in the error term that underlies the CLRM will be violated. Ø We experience autocorrelation when Topic Nine Serial Correlation

1. Introduction Ø Sometimes the term autocorrelation is used interchangeably. Ø However, some authors prefer to distinguish between them. Ø For example, Tintner defines autocorrelation as ‘lag correlation of a given series within itself, lagged by a number of times units’ whereas serial correlation is the ‘lag correlation between two different series’. Ø We will use both term simultaneously in this lecture. Topic Nine Serial Correlation

1. Introduction Ø There are different types of serial correlation. With first-order serial correlation, errors in one time period are correlated directly with errors in the ensuing time period. Ø With positive serial correlation, errors in one time period are positively correlated with errors in the next time period. Topic Nine Serial Correlation

2. Causes of Autocorrelation 1. Inertia - Macroeconomics data experience cycles/business cycles. 2. Specification Bias- Excluded variable Ø Appropriate equation: Ø Estimated equation Ø Estimating the second equation implies Topic Nine Serial Correlation

2. Causes of Autocorrelation 3. Specification Bias- Incorrect Functional Form Topic Nine Serial Correlation

2. Causes of Autocorrelation 4. Cobweb Phenomenon Ø In agricultural market, the supply reacts to price with a lag of one time period because supply decisions take time to implement. This is known as the cobweb phenomenon. Ø Thus, at the beginning of this year’s planting of crops, farmers are influenced by the price prevailing last year. Topic Nine Serial Correlation

2. Causes of Autocorrelation 5. Lags Ø The above equation is known as autoregression because one of the explanatory variables is the lagged value of the dependent variable. Ø If you neglect the lagged the resulting error term will reflect a systematic pattern due to the influence of lagged consumption on current consumption. Topic Nine Serial Correlation

2. Causes of Autocorrelation 6. Data Manipulation Ø This equation is known as the first difference form and dynamic regression model. The previous equation is known as the level form. Ø Note that the error term in the first equation is not autocorrelated but it can be shown that the error term in the first difference form is autocorrelated. Topic Nine Serial Correlation

2. Causes of Autocorrelation 6. Nonstationarity Ø When dealing with time series data, we should check whether the given time series is stationary. Ø A time series is stationary if its characteristics (e. g. mean, variance and covariance) are time variant; that is, they do not change over time. Ø If that is not the case, we have a nonstationary time series. Topic Nine Serial Correlation

3. OLS Estimation Ø Assume that the error term can be modeled as follows: Ø is known as the coefficient of autocovariance and the error term satisfies the OLS assumption. Ø This Scheme is known as an Autoregressive (AR(1))process Topic Nine Serial Correlation

3. Topic Nine OLS Estimation Serial Correlation

4. BLUE Estimator Ø Under the AR (1) process, the BLUE estimator of β 2 is given by the following expression. Topic Nine Serial Correlation

4. BLUE Estimator Ø The Gauss Theorem provides only the sufficient condition for OLS to be BLUE. Ø The necessary conditions for OLS to be BLUE are given by Krushkal’s theorem. Ø Therefore, in some cases, it can happen that OLS is BLUE despite autocorrelation. But such cases are very rare. Topic Nine Serial Correlation

5. Consequences of Using OLS Ø OLS Estimation Allowing for Autocorrelation Ø As noted, the estimator is no more not BLUE, and even if we use the variance, the confidence intervals derived from there are likely to be wider than those based on the GLS procedure. Ø Hypothesis testing: we are likely to declare a coefficient statistically insignificant even though in fact it may be. Topic Nine Ø One should use GLS and not OLS. Serial Correlation

5. Consequences of Using OLS Ø OLS Estimation Disregarding Autocorrelation Ø The estimated variance of the error is likely to overestimate the true variance Ø Over estimate R-square Ø Therefore, the usual t and F tests of significance are no longer valid, and if applied, are likely to give seriously misleading conclusions about the statistical significance of the estimated regression coefficients. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Ø Graphical Method There are various ways of examining the residuals. The time sequence plot can be produced. Alternatively, we can plot the standardized residuals against time. Ø The standardized residuals is simply the residuals divided by the standard error of the regression. Ø If the actual and standard plot shows a pattern, then the errors may not be random. Ø We can also plot the error term with its first lag. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø The Runs Test- Consider a list of estimated error term, the errors term can be positive or negative. In the following sequence, there are three runs. Ø (─ ─ ─ ) ( + + + + ) (─ ─ ───── ) Ø A run is defined as uninterrupted sequence of one symbol or attribute, such as + or -. Ø The length of the run is defined as the number of element in it. The above sequence as three runs, the first run is 6 minuses, the second one has 13 pluses and the last one has 11 runs. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø The Runs Test- Consider a list of estimated error term, the errors term can be positive or negative. In the following sequence, there are three runs. Ø (─ ─ ─ ) ( + + + + ) (─ ─ ───── ) Ø A run is defined as uninterrupted sequence of one symbol or attribute, such as + or -. Ø The length of the run is defined as the number of element in it. The above sequence as three runs, the first run is 6 minuses, the second one has 13 pluses and the last one has 11 runs. Topic Nine Serial Correlation

5. Ø Ø Ø Detecting Autocorrelation Define N: total number of observations N 1: number of + symbols (i. e. + residuals) N 2: number of ─ symbols (i. e. ─ residuals) R: number of runs Assuming that the N 1 >10 and N 2 >10, then the number of runs is normally distributed with: Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Then, Ø If the null hypothesis of randomness is sustainable, following the properties of the normal distribution, we should expect that Ø Prob [E(R) – 1. 96 R ≤ E(R) – 1. 96 R] Ø Hypothesis: do not reject the null hypothesis of randomness with 95% confidence if R, the number of runs, lies in the preceding confidence interval; reject otherwise Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø The Durbin Watson Test Ø It is simply the ratio of the sum of squared differences in successive residuals to the RSS. Ø The number of observation is n-1 as one observation is lost in taking successive differences. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø A great advantage of the Durbin Watson test is that based on the estimated residuals. It is based on the following assumptions: 1. The regression model includes the intercept term. 2. The explanatory variables are nonstochastic, or fixed in repeated sampling. 3. The disturbances are generated by the first order autoregressive scheme. Topic Nine Serial Correlation

5. Detecting Autocorrelation 4. The error term is assumed to be normally distributed. 5. The regression model does not include the lagged values of the dependent an explanatory variables. 6. There are no missing values in the data. Ø Durbin-Watson have derived a lower bound d. L and an upper bound d. U such that if the computed d lies outside these critical values, a decision can be made regarding the presence of positive or negative serial correlation. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Where Ø But since -1 ≤ ≤ 1, this implies that 0 ≤ d ≤ 4. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø If the statistic lies near the value 2, there is no serial correlation. Ø But if the statistic lies in the vicinity of 0, there is positive serial correlation. Ø The closer the d is to zero, the greater the evidence of positive serial correlation. Ø If it lies in the vicinity of 4, there is evidence of negative serial correlation Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø If it lies between d. L and d. U / 4 –d. L and 4 – d. U, then we are in the zone of indecision. Ø The mechanics of the Durbin-Watson test are as follows: Ø Run the OLS regression and obtain the residuals Ø Compute d Ø For the given sample size and given number of explanatory variables, find out the critical d. L and d. U. Ø Follow the decisions rule Topic Nine Serial Correlation

5. Ø Ø Detecting Autocorrelation Use Modified d test if d lies in the zone in the of indecision. Given the level of significance , Ho: = 0 versus H 1: > 0, reject Ho at level if d < d. U. That is there is statistically significant evidence of positive autocorrelation. Ho : = 0 versus H 1 : < 0, reject Ho at level if 4 - d < d. U. That is there is statistically significant evidence of negative autocorrelation. Ho : = 0 versus H 1 : ≠ 0, reject Ho at 2 level if d < d. U and 4 - d < d. U. That is there is statistically significant evidence of either positive or negative autocorrelation. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø The Breusch – Godfrey Ø The BG test, also known as the LM test, is a general test for autocorrelation in the sense that it allows for 1. nonstochastic regressors such as the lagged values of the regressand; 2. higher-order autoregressive schemes such as AR(1), AR (2)etc. ; and 3. simple or higher-order moving averages of white noise error terms. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø The Breusch – Godfrey Ø The BG test, also known as the LM test, is a general test for autocorrelation in the sense that it allows for 1. nonstochastic regressors such as the lagged values of the regressand; 2. higher-order autoregressive schemes such as AR(1), AR (2)etc. ; and 3. simple or higher-order moving averages of white noise error terms. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Consider the following model: Ø Estimate the regression using OLS Ø Run the following regression and obtained the R-square Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø If the sample size is large, Breusch and Godfrey have shown that (n – p) R 2 follow a chi-square Ø If (n – p) R 2 exceeds the critical value at the chosen level of significance, we reject the null hypothesis, in which case at least one rho is statistically different from zero. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Point to note: Ø The regressors included in the regression model may contain lagged values of the regressand Y. In DW, this is not allowed. Ø The BG test is applicable even if the disturbances follow a pth-order moving averages (MA) process, that is ut is integrated as follows: Ø A drawback of the BG test is that the value of p, the length of the lag cannot be specified as a priori. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Model Misspecification vs. Pure Autocorrelation Ø It is important to find out whether autocorrelation is pure autocorrelation and not the result of mis-specification of the model. Ø Suppose that the Durbin Watson test of a given regression model (wage-productivity) reveals a value of 0. 1229. This indicates positive autocorrelation Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø However, could this correlation have arisen because the model was not correctly specified? Ø Time series model do exhibit trend, so add a trend variable in the equation. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø The Method of GLS Ø There are two cases when (1) is known and (2) is not known Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø When is known Ø If the regression holds at time t, it should hold at time t-1, i. e. Ø Multiplying the second equation by gives Ø Subtracting (3) from (1) gives Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø The equation can be Ø The error term satisfies all the OLS assumptions Ø Thus we can apply OLS to the transformed variables Y* and X* and obtain estimation with all the optimum properties, namely BLUE Ø In effect, running this equation is the same as using the GLS. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø When is unknown, there are many ways to estimate it. Ø Assume that = +1 the generalized difference equation reduces to the first difference equation Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø The first difference transformation may be appropriate if the coefficient of autocorrelation is very high, say in excess of 0. 8, or the Durbin. Watson d is quite low. Ø Maddala has proposed this rough rule of thumb: Ø Use the first difference form whenever d< R 2. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø There are many interesting features of the first difference equation Ø There is no intercept regression in it. Thus, you have to use the regression through the origin routine Ø If however by mistake one includes an intercept term, then the original model has a trend in it. Ø Thus, by including the intercept, one is testing the presence of a trend in the equation. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Another interesting feature relates to the stationarity property. Ø When =1, the error term, ut, is nonstationary, for the variances and covariances become infinite. Ø When =1, the first differenced ut becomes stationary, as it is equal to q white noise error term. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Based on Durbin-Watson d statistic Ø From the Durbin – Watson Statistics, we know that Ø In reasonably large samples one can obtain rho from this equation and use it to transform the data as shown in the GLS. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Based on Durbin-Watson d statistic Ø From the Durbin – Watson Statistics, we know that Ø In reasonably large samples one can obtain rho from this equation and use it to transform the data as shown in the GLS. Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Based on the error terms Ø Estimate the following equation Topic Nine Serial Correlation

5. Detecting Autocorrelation Ø Iterative Procedure Ø We can estimate rho by successive approximation, starting with some initial value of rho. Ø the Cochran-Orcutt iterative procedure, the Cochran. Orcutt two-step procedure, the Durbin-Watson twostep procedure and the Hildreth-Lu scanning or search procedure. Ø The most popular one is the Cochran-Orcutt iterative procedure. Topic Nine Serial Correlation