It is hard to test macroeconomic theories because it is not possible to do controlled experiments to test them. When a scientific theory is to be tested, it is necessary to do a controlled experiment to test it. This is possible in a laboratory, but not in the real economic world.
Let us say, for example, that we want to find out whether it would be good for the US economy to allow the major automakers to fail rather than bailing them out (as was done in 2008 and 2009). To really answer this for certain, we would have to have history happen twice. We would need to have one history in which we bailed out the automakers. Then we would have to go back, run history again, and let the automakers fail. We would have to make sure that nothing else changed; the only change would be that we allowed the automakers to fail. Obviously, it is not possible to run history back and try things again.
Alternatively, we could just look at what actually happened. We know that we bailed out the auto industry. We could just look and see if our economy improved after the bailout. That should be a test of our theory. But the problem is that many other things happened that could have affected the economy. The bailout mattered, but so did things like economic crises in Europe. We would have a very hard time knowing whether the bailout affected our economy or whether the economy was affected more by a variety of other factors.
For these reasons, it is very hard to actually test macroeconomic theories.