Questions and Answers
-
Answer added in Socio-Economics2 Relationship analysis between social conflict and economic inequalityBy Sri Rosmardiyah · Sekolah Tinggi Ilmu StatistikPandelis Mitsis · University of CyprusYou don't give enough information for anyone to help you. "You found 3 theories that support your research". What do you mean by that? You made estim... [more]You don't give enough information for anyone to help you. "You found 3 theories that support your research". What do you mean by that? You made estimations and the 3 theories have predictions that agree with your results? In order for anyone to suggest to you an (econometric) method, you should first state the nature of your data (cross-section, panel, time series?).Following
-
Answer added in Applied Econometrics9 Can I make a regression model with the whole population?By Lucia Mauro · Universidad Nacional de Mar del PlataPandelis Mitsis · University of CyprusLucia, even if your number of observations is low (as indicated by previous users), I would suggest you go and run your regression, since those obser... [more]Lucia, even if your number of observations is low (as indicated by previous users), I would suggest you go and run your regression, since those observations are 83% of the whole population. You don;t state your variable of interest (e.g. profit of the firms?), but as again indicated above, you should choose wisely your independent variables due to the degrees of freedom (=number of observations - number of independent variables). Even if the regression will give result even with 24 ind. variables, I would not suggest more than 10. Methods for limiting regressons to the least number: a) Include all of them and then remove the ones with p-values higher than 0.10, 0.05 or 0.01 (depending on how strict you want to be). b) Make auxiliary regressions with one ind. variable at a time and include in the "final" regression only the ones with a p-value higher than 0.10 or 0.05 or 0.01. Note that in both cases the p-values of each variable may change from regression to regression for many reasons. Finally, as other users indicated, make sure that you make some mis-specification tests in your "final" regression. For a cross-section regression (as the one you suggest), I would check the residuals for heteroscedasticity (White test), Normality (Jarque-Bera test), Linearity (RESET test) or Parameter Stability (CUSUM test). These tests also have a p-value, which you may use to reach a conclusion about the model. When a test does not have a positive result, I woudl suggest you mention that the small number of obs may had sth to do with that. The small no. of obs. is the reason I do not propose a unit root as well. That is it, actually. That isFollowing
-
Answer added in Econometrics32 Johansen Cointegration to test the long run relationship between exchange rate and macroeconomic fundamentals.By Sujan Ghosh · Bangladesh Institute of Bank ManagementPandelis Mitsis · University of CyprusThank you Dr. Pal for your kind words. Ivette Luna, Ronald Kumar and Sujan Kumar Ghosh, you can find an empirical application of a thorough specific... [more]Thank you Dr. Pal for your kind words. Ivette Luna, Ronald Kumar and Sujan Kumar Ghosh, you can find an empirical application of a thorough specification of a VECM (in economic variables) in the following link: http://www.ucy.ac.cy/data/ecorece/DOA05-061.pdfFollowing
-
Answer added in Econometrics32 Johansen Cointegration to test the long run relationship between exchange rate and macroeconomic fundamentals.By Sujan Ghosh · Bangladesh Institute of Bank ManagementPandelis Mitsis · University of CyprusHi Sujan, Sorry if I repeat any of the topics above, but I don't have time to read them all. What it stroke me from your question is that you imply t... [more]Hi Sujan, Sorry if I repeat any of the topics above, but I don't have time to read them all. What it stroke me from your question is that you imply that the co-integration test and/or the VECM you estimated were the first and not last step of your modelling. In other words, first you check your model for mis-specification and then you use it for inference. Did you perform univariate unit root tests to your variables? Did you use AIC and SIC to determine the optimal lag length of your model? Did you estimate a VAR before the VECM? Did you check it for auto-correlation, non-Normality, heteroskedasticity or structural breaks or non-linearity? Did you test the co integrating rank? E.g. if is not 1 then maybe you don't have a stable long-run relationship. Granger-causality tests? I.e. to test whether your endogenous variables are indeed endogenous. If you performed all these then I apologize for this interrogation and I propose you the following in order to make your result more robust: a) Impulse/Response Analysis b) Variance Decomposition Analysis c) Out-of-sample or within-sample forecasts Good luckFollowing
-
Answer added in Econometric Analysis12 Modelling impact of money supply and oil price and exchange rate on trade balance?By Kushneel Prakash · University of the South PacificPandelis Mitsis · University of CyprusIn addition to the feedback you already received, I would say: (1) The fact that AIC suggests only 1 lag is very good, since you have only 45 observa... [more]In addition to the feedback you already received, I would say: (1) The fact that AIC suggests only 1 lag is very good, since you have only 45 observations and more lags will reduce the degrees of freedom of your estimations. (2) Since OIL is stationary then there is no reason to put it in first differences, unless if you want it to be consistent with the remaining variables. (3) VERY crucial for me: In your first output results you have all 4 variables in your VAR stationary (since they are in first differences) then how come you are finding co-integration? Econometrics say that co-integration may exist ONLY between integrated (=non-stationary) variables. So, I would suggest you work on the second outcome, if you wanna examine a long-run relationship (co-integrating vector). Have you tested the VAR model for co integration before you continued to VECM? (4) If the trend is significant in the ADF tests I would suggest you add it to your models. (5) I would suggest before continuing to Granger causality tests, Impulse Response or Variance Decomposition Analysis, to examine as an alternative to the second output: A trivatiate VAR (with LTR, LCPI, LREER) with LOIL and linear trend as an exogenous variable. Good luck.Following
-
Answer added in minimum wage6 What is the most appropriate minimum wage variable?By Pandelis Mitsis · University of CyprusPandelis Mitsis · University of CyprusThank you for your answer Joshua Zambrano. However, you must notice that in estimating models as the one I suggest, one may avoid the aforementioned... [more]Thank you for your answer Joshua Zambrano. However, you must notice that in estimating models as the one I suggest, one may avoid the aforementioned exchange rate/currency factor by using the ratio of the minimum wage to the average wage, or using the percentage changes of the real minimum wage. Actually, the minimum wage/mean wage ratio (known also as Kaitz Index) reveal another aspect when examining this topic: you should always take into consideration the median or mean wage. E.g. an effective minimum wage is expected to be at least equal to the national median wage.Following
-
Answer added in Econometric Analysis4 When dealing with time series annual or quarterly data. What logical steps should be followed when trying to find out some long run and short run relationships among your variables?By Kushneel Prakash · University of the South PacificPandelis Mitsis · University of CyprusSorry for the long answer, but if you want to have robust results you have to follow these steps: A) First determine whether your series an stationar... [more]Sorry for the long answer, but if you want to have robust results you have to follow these steps: A) First determine whether your series an stationary or not - There are many unit root tests available in E VIEWS, I recommended ADF. B) Then accordingly with the unit root test results, you may estimate a vector auto-regression (VAR) model (also easy in E-VIEWS), using the number of lags indicated in the previous tests: Case 1: Both series are stationary (i.e. there are no unit roots). Then there maybe ONLY a short-run relationship between your variables. You estimate a simple VAR model. The coefficients of the lagged values will indicate the said short tun relationship. E.g. if the coefficients of the lagged values of the first variable are statistically insignificant in the equation of the other variable, that indicates NO short-run relationship - And vice versa. Case 2: Both series are NON-stationary (i.e. they both have unit roots). Then there maybe short-run AND long-run relationships between your variables. In order to examine the existence of a short-run relationship you do exactly as above (i.e. check the relevant coefficients in the VAR model). In order to check the existence of a long-run relationship then you FIRST perform a Co-integration test (also available in E-Views). If, you have 1 co-integrating vector THEN you go and estimate a vector error correction model (VECM). The estimated co-integration vector IS your long-run relationship (i.e. check the signs and statistical significance of the coefficient(s)). If the co-integration test indicates 0 or 2 co-integrating vectors, then there is NOTa stable long-run relationship between your variables and you emphasize only the VAR estimation results. Case 3: Only ONE series is stationary. Then you use the first-difference of the variable that is NOT stationary (and the stationary variable as it is) and then you do exactly the same as in Case 1. (Then there maybe ONLY a short-run relationship between your variables. )Following
-
Answer added in minimum wage6 What is the most appropriate minimum wage variable?By Pandelis Mitsis · University of CyprusPandelis Mitsis · University of CyprusThanks again, Yes, I refer to the Hirsch and McPherson , "Minimum Wage Effects on Teen Employment Transitions and Wage Growth". However, as they stat... [more]Thanks again, Yes, I refer to the Hirsch and McPherson , "Minimum Wage Effects on Teen Employment Transitions and Wage Growth". However, as they state in the end of page 7-start of page 8, the minimum wage variable they use varies across states AND areas, as much as across years. From what you way though perhaps I should unify the 5 cross-section samples into a an unmatched panel data set. However, I am afraid that the minimum wage variable would then be perfectly correlated with the year dummies (or the time trend) and thus dropped out from the regressions. Is that possible?Following
-
Answer added in minimum wage6 What is the most appropriate minimum wage variable?By Pandelis Mitsis · University of CyprusPandelis Mitsis · University of CyprusThanks for your reply. I actually have 5 cross-section data sets, each 7 years apart, BUT, are not matched, .i.e. I cannot examine the effect of the m... [more]Thanks for your reply. I actually have 5 cross-section data sets, each 7 years apart, BUT, are not matched, .i.e. I cannot examine the effect of the minimum wage across time in each individual. What I am thinking is to measure the minimum wage in a way that I have cross variation in terms of occupations. E.g. for every occupational group to calculate the ratio to the minimum wage to the average wage of the given occupation. I found a study that does this in order to create interstate variation of the minimum wage variable (where the minimum wage is identical). It uses the average wage (unskilled workers) in each area in the state as the denominator in the minimum wage variable (Hirsch and McPherson, 1999). What is your opinion on that?Following
-
Question asked in minimum wage6 What is the most appropriate minimum wage variable?I am using a Probit model and cross-section data to determine the effect of minimum wage in the probability of one being employed/unemployed. The min... [more]I am using a Probit model and cross-section data to determine the effect of minimum wage in the probability of one being employed/unemployed. The minimum wage is the SAME in all the sample (no cross-sectional variation) so I cannot use it as a regressor. The difference of the minimum wage from the persons' current wage cannot be calculated for the unemployed, unless if i use the wage of his/her last employment. Is this appropriate and if not then what measure would be appropriate, given the aim of my study?By Pandelis Mitsis · University of CyprusFollowing
-
Answer added in Economics12 How do I analyse transaction costs, bargaining power, institutional change, barriers of market entry and enforcement cost quantitatively?By Elias Danyi Kuusaana · University for Development StudiesPandelis Mitsis · University of CyprusI have only one topic for you: search and matching models in the Labour Market. Pissarides and Mortensen recently won a Nobel for this. For an idea ho... [more]I have only one topic for you: search and matching models in the Labour Market. Pissarides and Mortensen recently won a Nobel for this. For an idea how the factors you mentioned may be estimated in a search and matching framework see: Flinn, C. J. (2006), "Minimum Wage Effects on Labor Market Outcomes under Search, Matching, and Endogenous Contact Rates," Econometrica, Vol. 74(4), pp. 1013-1062.Following
-
Answer added in Economic Development9 What economic model can I use to explain the causes of informal payments in health care systems, mostly in FSU countries?By Anna Vardanyants · Universität TrierPandelis Mitsis · University of CyprusHi. It all depends on the available data. I.e. on how you intend to measure your key variables - informal payments in the healthcare systems. What ki... [more]Hi. It all depends on the available data. I.e. on how you intend to measure your key variables - informal payments in the healthcare systems. What kind of available data you have - time series, panel data or cross-section? The data will "tell" you which model to use.Following

