This is true. There will (almost) always be outliers due to confounding variables. Human society is really, really complex and responds to theories about itself so there are a lot of factors going on, many are not constant. However, that should not stop us from trying to discover what is most likely to occur.
For example, looking for a correlation between
A) How "corrupt" is a country?
B) How "Socialist" is a country?
Of course it is also difficult (I think not impossible) to define and measure "corruption" and "socialist" to some practical degree of precision. But if this hurdle is overcome, and these measurements are taken year after year, and we see that as a country becomes more socialist it becomes more corrupt (and/or vice versa), this would be scientific evidence that the amount of socialism is related somehow to the amount of corruption. Note that I am not using historical examples here. In other words, the hypothesis that degree of socialism is unrelated to corruption (during the period of time measured) will have been falsified. This seems to be pretty clear cut science to me.
Note that whether or not this finding will still apply 100 years later after various cultural, geopolitical, and technological changes have occurred would still be an open question. To get good enough data to answer that, these types of experiments would probably need to be done on the scale of millenia.