Dear fellow statalists,
This is my first post on the forum, so apologies if my query is trivial. At this I will make first a few clarification, in my attempt to ensure that the question in well understood.
Say I have a set of earnings data (across a set of firms for a 5 year window). Looking at the first year, I calculate the median earnings value across all firms. I proceed to create a new variable (net_earnings) equal to the difference between the actual and the year median. Afterwards, I allocate the firms into deciles according to net earnings.
I repeat this for every year and theoretically I should obtain something like the graph below, where the differences to the mean become smaller as time passes. In other words, net_earnings would be tending towards a mean value.
Is there a statistically reliable way to distinguish between fast vs slow reverting process ?
A fast process in my mind is one that can take place in one period as oppose to five periods. In addition, there is also the element where a firm experiences such a high level of reversion that it moves from one decile to another (e.g. d10 -> d8).
My overall aim would be to put a label (slow/normal/fast) for every firm year. Say if my overall time interval is 30 years, I intend to utilise 5 year rolling windows to label the degree of reversion a firm finds itself in that particular year.
I would be very grateful for any advice. I have problems in conceptually grasping this relationship and implementing it with Stata.
Best regards,
Andrei
This is my first post on the forum, so apologies if my query is trivial. At this I will make first a few clarification, in my attempt to ensure that the question in well understood.
Say I have a set of earnings data (across a set of firms for a 5 year window). Looking at the first year, I calculate the median earnings value across all firms. I proceed to create a new variable (net_earnings) equal to the difference between the actual and the year median. Afterwards, I allocate the firms into deciles according to net earnings.
I repeat this for every year and theoretically I should obtain something like the graph below, where the differences to the mean become smaller as time passes. In other words, net_earnings would be tending towards a mean value.
Is there a statistically reliable way to distinguish between fast vs slow reverting process ?
A fast process in my mind is one that can take place in one period as oppose to five periods. In addition, there is also the element where a firm experiences such a high level of reversion that it moves from one decile to another (e.g. d10 -> d8).
My overall aim would be to put a label (slow/normal/fast) for every firm year. Say if my overall time interval is 30 years, I intend to utilise 5 year rolling windows to label the degree of reversion a firm finds itself in that particular year.
I would be very grateful for any advice. I have problems in conceptually grasping this relationship and implementing it with Stata.
Best regards,
Andrei
