Hello,

I have a problem with chosing the right controls for my model and hope someone can help me along .

With the model I want to explain savings, human capital and labour supply by the time of the demographic Transition (DT). DT is the variable of interest. The model is the same for all three:

dependent_2010 = ß_0 + ß_1*dependent_1990 + y1*DT + y2*DT² + c*Control , with subscripts i for the variables


I have already chosen a proxy for urbanity and a dummy for war during the investigation period 1990-2010 as control variables.

Question 1:Do I have to expect that urbanity/war is both correlated with an independent variable AND affects the dependent variable - or is it enough to assume that urbanity/war affects the dependent variable WITHOUT correlation to any independent variable?? -- In case that correlation with an independent variable is needed: My model includes one lagged value. If I assume that a control variable affects the dependent variable, I simultaneously assume it affects one of the "independent" variables, the lagged value, too. However, I have a feeling this would not force me to include such acontrol variable in the model, does it? I hope this makes sense.

Question 2: I have though about including the level of income in 1990 but would this be purposeful? Ecomomic models usually explain income as a function of savings, human capital and labour supply - so explaining these variables with income would be misleading, although it sounds logical that poor nations save less or can only afford little education. Should I exclude variables where reverse causality could occur?

Question 3: To control for life expectancy would be reasonable in my sense because expecting a longer life, people could tend to save more, work more and get more education. Now, this is tricky for me: The demographic transition (DT) is initiated by falling mortality rates and by this manner, also by rising life expectancy. That means life expectancy and DT must be correlated. Is it still okay to include life expectancy it in the model? I fear that because DT is a result of
life exp., I could erase potential effects of DT on the dependent variable?

Off Topic- Question 4: After I receive the estimation results and, say, find significant effects of DT on the dependent variable. What phrases am I "allowed" to state? I surely can't say, 'The result is that DT causes the dependent variable to rise/fall.' Is this really all I can say: 'We cannot reject that the effect of DT on the dependent variable is non-existent'?


I apologize as this is no direct question about Stata, but the help received on this forum is very valuable and I don't know where else to ask.
As always, thank you!!