Hi dear Scholars and everyone,

I hope all of you are doing well.

I have a question. If you could help me by sharing your knowledge to answer my questions, I shall be so grateful in advance. I need to test my quarterly data which is to estimate earnings persistence.

I use variations of the seasonally differenced quarterly earnings model (5) (ARI) proposed by Foster paper where the model is shown as follow:

Model : Q t + 1 = β 0 + β 1 Q t + β 2 Q t+3+ β 3 Q t+4+ ε

Where Qt+1refers to earnings quarterly on one a head period. Fostor used Q t+3and Q t+4 to capture the controlling for the seasonal effects. The lagged value ( Q t ) of the dependent variable by its coefficient interprets the level of relation between dependent variable (Qt+1) and its first lag (Q t) (persistence). allowing the coefficients to vary over the quarterly components. I examine the relation among the quarterly components and one-period-ahead earnings Qt+1 by decomposing the Qt terms into the respective lagged terms.

I have divided my sample into two sub periods which are financial crisis period and non-financial crisis period and used the model above. I want to examine the effect of the crisis on quarterly earnings persistence and find if there is difference among (two different economic conditions). My question to you is, do I need to remove the two lags (Q t+3 and Q t+4) from the model in case of testing the data in a period of financial crisis? In other words, do you think remaining of these lags (Q t+3 and Q t+4) could affect the results in terms of examining earnings persistence in a period of financial crisis? Or is it not?

Or, do you have information that can suggest an alternative model could be appropriate for my case to estimate persistence in quarterly earnings in accounting. It is appropriate if it can be used in accounting studies.

Also, could I just use the autoregressive distributed lag (AR) model or ordinary least square (OLS) model in accounting study, and that is by using only the first lag (Qt ) difference as independent variable in the model and some other independent variables in case we need it for the study. I hope you can answer me. Kind regards