Designing a survey?

I am designing a survey to give out at work to find out why the turn over rate has increased so much in the last two years. What type of variables should I inquire about. My hypothesis is: Employees who feel underappreciated at work tend to have a poor work attitude, thus move on to other companies.

Designing a survey?
Job satisfaction. What place the job held in their live (like, was it a career move, a transistion, something to pay the bills). Burn out/stress. If they feel underappreciated/what would make employees feel more appreciated. General work ethic.
Reply:tyrant bosses,low pay,overtime,employee rivalry,bad equipment,induced stress,even personal problems at home,not to mention the meth problem has become epidemic and users cannot function even though they think they can.GL
Reply:Some possible survey variables:





Age and sex: younger people may leave more often than older people





Job title and years of experience at the company and in the industry: more experienced people tend to stay, non-managers might leave more frequently





Company benefits and salary, such as 401(k), medical insurance: companies with poor benefits may promote greater turnover





Industry and industry competition/competitors: certain industries and companies experience greater turnover due to healthier competitors





Satisfaction rating of current job and perception of growth opportunities in the company: few internal promotion opportunities make people leave





These are just a few, there are probably more. "Underappreciation" can be a difficult thing to measure since it's impacted by so many things - what if the benefits package is poor and so they feel "underappreciated"? What if their boss is poor and they feel "underappreciated"? That feeling can be generated by a number of different variables.
Reply:I'm not a psychologyist; thus, I might not fully understand your problem. However. It sounds as if you want to 'probe' whether your hypothesis is correct. It calls for a 'Contingency Table' kind of analysis. You should look further on that subject on any basic statistics book.


Now, you must survey both godd-attitude people, as well as bad attitude. You should survey people who have stayed, as well as people who have left.


You can draw and fill a table like this:





.............Stay..|.Left.|.Total.|


----------------------------------


Good.|............|........|.............


----------------------------------


Bad ..|............|........|..........|


----------------------------------


Total.|............|........|......... |





If your hypothesis is WRONG , good and bad should be distributed EVENLY (proportionally) among left and stayed:


E(Good-Left) = Total Good x Left/Total


E(Bad-Left)=Total Bad x Left / Total





Then you compute for each entry expected minus observed, squared.


The sum of all those distributes as a Chi square with N minus 2 degrees of freedom.


If the computed sum is HIGHER than the correspondant CHi square P-value, the your hypothesis is true with 1-p probability.


Hope it helps