Formal tests of volatility in STATA17

[Eddie Brighton]

Dear Statlist:

I have longitudinal data on employee data across multiple years and organizations. I've been making a series of line graphs to visualize the trends in employee size across time by organizations. These are intuitive, but I wonder if there are formal tests/commands that let us test volatility. My goal is to see which organizations had the greatest fluctuation in terms of employee size (& salary) over the years. Thanks in advance, and here's an example of my data where "agy" is organization and "adjbasicpay" salary.

* Example generated by -dataex-. For more info, type help dataex
clear
input float year str2 agy long adjbasicpay
1973 "AG" 18080
1973 "AG" 36000
1973 "AG" 34992
1973 "AG" 26671
1973 "AG" 36000
1973 "AG" 33915
1973 "AG" 12056
1973 "AG" 24247
1973 "AG" 30147
1973 "AG" 28267
1973 "AG" 34971
1973 "AG" 36000
1973 "AG"  9235
1973 "AG" 11961
1973 "AG"     .
1973 "AG" 15009
1973 "AG" 33899
1973 "AG"  9144
1973 "AG" 36000
1973 "AG" 30486
1973 "AG" 30147
1973 "AG" 16609
1973 "AG" 12634
1973 "AG" 34074
1973 "AG"  9969
1973 "AG"  9874
1973 "AG" 24247
1973 "AG" 21014
1973 "AG" 40000
1973 "AG"  7198
1973 "AG" 14053
1973 "AG" 24247
1973 "AG"     .
1973 "AG" 15331
1973 "AG" 36000
1973 "AG" 10002
1973 "AG" 31089
1973 "AG" 14671
1973 "AG" 16609
1973 "AG" 26898
1973 "AG" 13379
1973 "AG" 36000
1973 "AG" 32973
1973 "AG" 17497
1973 "AG" 10528
1973 "AG" 11961
1973 "AG"  8757
1973 "AG" 15331
1973 "AG" 34971
1973 "AG" 38000
1973 "AG" 12373
1973 "AG" 24247
1973 "AG" 15609
1973 "AG" 12501
1973 "AG" 33177
1973 "AG" 12283
1973 "AG"  8591
1973 "AG" 17605
1973 "AG" 15649
1973 "AG" 28267
1973 "AG" 13687
1973 "AG" 31089
1973 "AG" 12283
1973 "AG" 31089
1973 "AG" 34971
1973 "AG" 21671
1973 "AG" 34965
1973 "AG" 13996
1973 "AG"  8299
1973 "AG" 19246
1973 "AG" 23858
1973 "AG" 10234
1973 "AG" 25055
1973 "AG" 13336
1973 "AG" 11297
1973 "AG" 12979
1973 "AG" 33899
1973 "AG"  8299
1973 "AG" 36000
1973 "AG" 12056
1973 "AG" 14603
1973 "AG" 12634
1973 "AG" 31383
1973 "AG" 14928
1973 "AG" 36000
1973 "AG"  9493
1973 "AG" 36000
1973 "AG" 38000
1973 "AG" 10860
1973 "AG" 28267
1973 "AG" 29205
1973 "AG" 34971
1973 "AG" 36000
1973 "AG" 36000
1973 "AG" 33177
1973 "AG" 16609
1973 "AG" 12985
1973 "AG" 23088
1973 "AG" 15945
1973 "AG" 15009
end

[Nick Cox]

A test requires a hypothesis, so what would the hypothesis be? Just asking for those who would know an answer given that.

[Eddie Brighton]

Hi Nick, I my hypothesis would be something like...
"Organization A's fluctuation in employee size is larger than Organization B, C, D, ..."
Or, I guess I'm just looking for an apprpriate indicator of volatility. For instance, generating a variable something like (# of employee in year n - # of employee in year n-1) / # of employee in year n. But have a strong sense that there's a better approach.

[Nick Cox]

That sounds to me a measurement problem, and the appropriate tool is a table or graph that shows organization volatility, however defined, in a direct manner.

[Eddie Brighton]

Makes sense, thanks!