inflows in flat neighboring countries affects the probability of flat tax change for their
neighbors.
All flat tax policy changes not only radically simplified taxation and “flattened” the
rates, but also reduced taxes considerably: on average, flat countries impose 30.5 percent
of top marginal individual tax rate, while non-flat do 38 percent. Because we can estimate
the effects of spatial lag in OLS regression model, we specify model D as the OLS regres-
sion and use the same variables to predict individual tax rates. Parameters that reduce
tax rates should be associated with those increasing the probability of flat tax adoption.
The estimate of the spatial-lag coefficient in the base model is statistically significantly
positive. The spatial multiplier, (I - ρW)-1 (Franzese & Hays 2006, 7-12) captures the
effects from one Eastern European country on others closest to it, from those — on their
neighboring countries, including back on that country. Multiplying (I - ρW)-1 by a 20x1
column-vector with 0 in all rows except for example, Lithuania, which receives a 1, pro-
duces a 20x1 vector that contains the estimated effects of a unit-shock (1 percent change
in taxation) in Estonia on the other 19 countries in their rows. We estimated standard
errors via statistical simulation. The resulting matrix provides the estimated effects of a
unit-shock to country i on tax policies in the other 19 countries j.
[TABLE THREE ABOUT HERE]
Table 3 reports the short-run spatial effects of marginal individual tax rates in Eastern
Europe. The first number in each cell is the immediate effect of a unit-increase (1 percent)
in the column country’s top marginal individual tax rate on other Eastern European
nations (rows). The results suggest that a change in taxation levels in a “flat” country
leads to a reduction in taxation in other countries. For example, decrease in taxation in
Estonia brings about immediate tax changes in Latvia, Lithuania, Ukraine and Russia.
26