Hello everybody,

I am trying to find out if there is a relationship between business tax increases and change in total assets (growth_assets_w). I want to use a bar chart to illustrate my diff in diff approach and I am trying to add fixed year and industry effects to the bar chart following the approach of Florian Heider and Alexander Ljungqvis in their paper "As certain as debt and taxes: Estimating the tax sensitivity of leverage from state tax changes" (Paper: https://www.sciencedirect.com/scienc...4405X15000057?). Also, I do not understand why the bar of "No tax change" changes over the years.

The relevant part of the paper:

Array


Fig. 3. Annual changes in leverage around state tax increases and state tax cuts. The figures plot the average annual within-firm change in long-term leverage net of the contemporaneous leverage change in the firm׳s SIC4 industry (to remove the influence of time-varying changes in industry conditions or nation-wide variation in business conditions that affect all industries simultaneously). The sample includes all firms experiencing a corporate income tax change in their home state over the period 1989–2011 and, as controls, all firms not subject to a tax change in their home state. Leverage changes are plotted for each year in a five-year window centered on the year a state increases or cuts its corporate income tax (year 0) for treated firms (striped bars) and controls (dotted bars). The difference between the two bars in a given year is the difference-in-difference estimate. The significance of t-tests (using standard errors clustered at the state level) of the null that the diff-in-diff is zero is indicated using asterisks. We use *** to denote significance at the 1% level.

I have the following dummies:
  • tIncreaseM2 = 1 if tax increase in T-2
  • tIncreaseM1 = 1 if tax increase in T-1
  • tIncrease = 1 if tax increase in T=0
  • tIncreaseP1 = 1 if tax increase in T+1
  • tIncreaseP2 = 1 if tax increase in T+2
That´s my code so far (ignoring industry and year fixed effects):


Code:
gen tIncreaseM2G = growth_assets_w if tIncreaseM2 == 1
gen tIncreaseM1G = growth_assets_w if tIncreaseM1 == 1
gen tIncreaseG = growth_assets_w if tIncrease == 1
gen tIncreaseP1G = growth_assets_w if tIncreaseP1 == 1
gen tIncreaseP2G = growth_assets_w if tIncreaseP2 == 1

graph bar (mean) tIncreaseM2G (mean) tIncreaseM1G (mean) tIncreaseG (mean) tIncreaseP1G (mean) tIncreaseP2G
Later I am using the following regression (with industry and year fixed effects):

Code:
reghdfe growth_assets_w tIncreaseP1 `firmSpecificControlVars2' `stateSpecificControlVars2', absorb(year industry_sic) vce(cluster industry_sic)

I would really appreciate any help!

Best regards
Andreas