Relationships change, why


Fortran must die
I should say ahead of time I have no theory for this, if I did I would use it.

We have a dependent variable that shows whether customers are or are not satisfied. The predictors is how satisfied they are with a variety of other factors like pay, the importance of their work, helping others etc.

Satisfaction with pay is extremely low now as it was six years ago the last time we ran this analysis. But, using the odds ratios to determine relative importance, pay had very limited impact on overall satisfaction 6 years, now it is the 2nd highest factor in terms of relative importance.

Is there any statistical way one can explore why this change occurred (some of the variables are new and some have been dropped from six years ago although most are the same).


TS Contributor
Did the variance of satisfaction with pay markedly increase?

Was satisfaction with pay highly correlated with the other variables 6 years ago, and is it now less related to other variables?

With kind regards



TS Contributor
Satisfaction is situational and will inherently change. If you have not read about the Kano model, I highly recommend doing so. I won't go through everything in the article, but want to point out the following:
  • People will typically express satisfaction/dissatisfaction only about performance needs. Basic needs will only drive comments about dissatisfaction.
  • Yesterday's performance needs become tomorrow's basic needs.
The economic situation can influence how pay is perceived. If the economy is doing great, there may be a general satisfaction with current pay, but if prices increase, satisfaction with pay would decrease.