Hi Valeska. In economic models the regression is usually something like Y= a+bX. Here the X's are key variables you think might explain the variation in Y (your dependent variable). The 'a' is simply a constant which shows you how far up the vertical axis you can move before you begin to understand the effect of X on Y (based on the coefficient b). Here the X is on the horizontal axis and the Y on the vertical axis. So leave the constant in the model and simply assume that the total effect of X on Y starts from the origin (0,0). Hope this helps? Cheers, Marc
Hi Valeska. In economic models the regression is usually something like Y= a+bX. Here the X's are key variables you think might explain the variation in Y (your dependent variable). The 'a' is simply a constant which shows you how far up the vertical axis you can move before you begin to understand the effect of X on Y (based on the coefficient b). Here the X is on the horizontal axis and the Y on the vertical axis. So leave the constant in the model and simply assume that the total effect of X on Y starts from the origin (0,0). Hope this helps? Cheers, Marc
Hi Garcia - you may need to ask yourself: what kind of model am I utilising? Is it descriptive, prescriptive or predictive? I think the significance of the constant/intercept is important if you have a predictive model (sometimes a prescriptive model). However, if the model you have is more of a descriptive one, you need not worry about the constant, as long as it was not dropped during the regression exercise. I hope that helps. David
hello David, well i contructed the model just to know how sensitive is the exchange rate(dollar versus local currency) in ADRs shares (ADR shares are traded in USA). i just need to calculate that and see how much affects the returns. i t
1. If the constant is important in your economic model (ex: you investigate the impact of output growth on private consumption, the constant is seen as the predetermined private consumtion in the case of zero disposal income), thus don't remove it
2. In the opposite case, you can remove it and then re-regress your model again to see the results and then analize it (ex: you investigate the evidence of taylor's rule of central bank, in which the target of monetary policy is potental output and zero gap in inflation so that the constant should be zero)
First of all, you must have read and done some ground work on your decision to have the constant in your model.
Therefore, it could be your sample that lead to insignificant of the said constant. I will humble suggest that you report it as insignificant and try to look for support in your context and setting that lead to the insignificant result.
Meanwhile, if there is the opportunity, test the same model, without the insignificant constant in a different setting.
Leave the constant and try to investigate some underlying factors that could explain the insignificant constant in your context. This factors you can use to justify your results and will also form the basis for further research.
Run the model without constant and see if the results are credible,however you need to provide empirical backup on the choice of such model as there can be underlying assumptions rendering the constant insignificant.
It depends on the interpretation you give the constant. Does it have a significant explanation in the model you have developed or is it just the intercept of the vertical axis? I agree with the response that you run the regression without the constant and consider the interpretation of the results.