Perhaps your antagonist will accept the following line of argument. Effects in the social order are the result of a multiplicity of causes. Theory is the only way to determine what effect a particular cause will have. Empirical correlation cannot do so.
For example, Paul Krugman was criticized that his advice that the government try massive fiscal and monetary stimulus to bring the economy out of the recent downturn had not worked by 2011. His response was not to concede that the correlation of massive stimulus and continuing depression demonstrate cause and effect, but to reply that the stimulus had not been big enough. But in doing so he was making a theoretical claim , not an empirical one, because he had no evidence to back him up.
If your antagonist will accept this line of argument, as Krugman implicitly does, then perhaps you can reason with him about taxes. And, as you say, the main point is that entrepreneurs in the market can make efficient production decisions using economic calculation while bureaucrats in the state cannot. So, economic progress depends on the extent to which taxing transfers command over resources out of private hands into the hands of the state. Interestingly, this has little to do with tax rates. The portion of GDP controlled by the government has stayed remarkably consistent since the early 1950s regardless of tax rates. Here are two charts: the first taxes and the second expenditures.
Since the early 1950s, federal government taxes as a percent of GDP have hovered between 15 and 20 percent.
Since the early 1950s, federal government spending as a percent of GDP has been between 18 and 24 percent.
The data are in Table 1.2 at the link below: