As far as I know ECM term should not be lower than -1, is not a good sign for your model. It implies that the process it not converging in the long run. ECM term must be between -1 and 0. You should check the long run relationship between your variables.
as for as i know there are two school of thouth , first it should be in betweeen 0-1 and the second thought is it coould be till -2. logically if we got ECT value-1 it implies that if there will be 100% conversion . but if ECT Comes -1.20 it indicates speed of conversion is 120% it implies dependent variable will come back before one year or whatever frequency you have.
When considering the ARDL (1,1) model, an assumption about the parameter connected with lagged dependent variable (alpha) needs to be made. We assume that the absolute value of the parameter is lower than 1. The ARDL (1,1) model can be presented as an ECM regression. In this model the speed adjustment parameter lambda=1-alpha . Therefore in ECM model lambda lies between -2 and 0.
If ECM parameter is between -2 and -1, This can suggest that the discrepancies between shocks and the trend are reduced in less than one year.
Article International Competitiveness of Czech Manufacturing: A Sect...
Problem with Error correction coefficient? - ResearchGate. Available from: https://www.researchgate.net/post/Problem_with_Error_correction_coefficient [accessed Nov 24, 2016].
Yes it can be less than -1. depending on the data frequency, it is interpreted as the proportion of the deviation from the long-term trend that is corrected in the first year, month, or quarter of the shock. however, you may also pay attention to the size of the ECM since a very small Error correction term signifies sluggishness in the system; in which case the workings of the market may take a longer time to achieve steady state. this becomes crucial especially if the shock tends to produce negative outcomes in the market which may necessitate government intervention of whatever form. for example in commodity market, demand for a commodity can negatively be hit by an adverse price shock. this will move demand down the equilibrium level. thus, it creates a problem where many people do not get access to good. if the demand system is very sluggish, it means the negative price shock will persist for a very long time, as well as the adverse effects it creates. in such a situation, the workings of the market may not be welfare-enhancing, and this may require some form of government intervention either fixing a price below the market level or providing subsidies to the most vulnerable section of the population. Hope this response prove useful.
The answer can be found here: https://faculty.washington.edu/ezivot/econ584/notes/cointegrationslides2.pdf
As suggested by Johansen (1995), with a simple example of a bivariate co-integrated VAR(1), a stability condition for an error-correction model is that the error-correction term must be strictly inferior to zero and strictly superior to -2. In effect, the error-correction model can be written as an AR(1) process for the disequilibrium error.
Full reference :
Likelihood-based Inference in Cointegrated Vector Autoregressive Models Advanced texts in econometrics, ISSN 1754-5765 Likelihood-based Inference in Cointegrated Vector Autoregressive Models, Søren Johansen Oxford scholarship online
The coefficient of ECM is also consistent with the rule of thumb which suggests that, the coefficient of error correction term (ECT) should be negative, less than one in its absolute value and significant.