These are the filters which I use to distinguish silly theories from credible ones.
1. Does it rely on a grand conspiracy?
This first filter eliminates theories about control by Illuminati, free masons, lizard people, and racist fantasies about a Jewish new world order. Conspiracies are tough to maintain because people like to talk. Even the threat of being killed won't stop death-bed confessions. It's just not believable that conspiracies will survive exposure, so theories that rely on this assumption are bunk.
2. Is there an absolute villain?
Closely tied to #1, some explanations rely on a class of super-villains or evil people who will stop at nothing to oppress others. An obvious pick would be the language of evangelical preachers, but also sometimes from Democrats claiming that the GOP is holding out on their plans for social betterment for the sole purpose of hurting working Americans.
Everyone is guided by some set of moral values and no one wants to feel like the bad guy. I think that generally everyone is trying to do the right thing, but disagree about the means. Even if we think 1% of the population is sociopathic, they'll have a hard time cooperating so the damage done will be localized. Postulating that everyone in the opposing camp is a psychopath is not credible.
3. Is it falsifiable?
Now the tests start to become more strenuous. Any social science theory must admit to some evidence which might disprove it, otherwise it has no way to be tested or compared against other theories. Belief in it becomes equivalent to dogma or faith.
For me, this part discredits the uber-government-spending brand of Keynesianism espoused by Paul Krugman. To summarize his view of history, when there was economic growth it was because of high government spending, and when there was stagnation it could have been growth if only the government had spent more. No historical circumstance could disprove this belief, so it can't be taken very seriously.
4. Do the assumptions about people make sense?
Economics in general is often received skeptically because it assumes people are rational. That is mostly based on a misunderstanding. Economic models rely on rationality in much the same way that a physics equations rely on zero friction or air resistance: the assumption simplifies analysis when fully realistic exposition is unnecessary, but the level of detail can be increased as necessary by adding in additional parameters.
Instead, this test bites against Utopian assumptions of a perfectly-improvable human nature. If progressive ideals could be realized, if only we tried hard enough, why haven't thousands of years of effort by religious leaders and philosophers gotten us there already? The answer must rely on some combination of #1 and #2 above. Theories of government which don't account for self-interested and imperfect human nature, such as Marxism or anarchy, are discarded.
5. Assuming the theory is true, what must also be true?
In other words, what are the theory's predictions? (If the theory makes no predictions, it's not a very interesting theory). Milton Friedman's famous 1953 paper, "The Methodology of Positive Economics", argued that all we should consider is a theory's predictive value, and not concern ourselves with the realism of its assumptions at all. Some take this as an extreme view, and perhaps Friedman's claim should not be followed dogmatically, but it's still a helpful insight in the social sciences.
For example, some theorists blame colonialism for the current poverty in Africa, Latin America, and Asia. But if this theory is true, we'd expect all colonized countries to have bad economies, and that isn't the case. The theory has to be revised. Daron Acemoglu has convincingly argued that the legal institutions put in place by the colonizers matters: Spanish imperialists were mostly looking to extract natural resources, so they didn't bother setting up legal institutions or property rights systems. The British had a more long-term view, which is why some of their former colonies (e.g. Hong Kong) are doing so well.
As another example, some strands of feminism argue that men, as a class, oppress women through a patriarchy. If this is true, how does one explain the great recent advances of women in voting rights, the workplace, no-fault divorce, and so on? Raquel Fernandez of NYU has an interesting paper which postulates that as economic development occurs, fathers' concern for their daughters leads them to support women's liberation. In a poor society, men benefit more from oppressing their wives, but as they get richer, they start to think about the effects of other men oppressing their daughters. This is a much better explanation for why gender norms change over time.
Out of many potential belief systems, I find these heuristics useful in narrowing down the number worth considering. Given the number of pseudo-scientific and anti-rational beliefs that still persist in the modern era, many people could benefit from a more skeptical world view.
A final quote from the ultimate conspiracy novel, Foucault's Pendulum, by Umberto Eco. It is
Not that the incredulous person doesn't believe in anything. It's just that he doesn't believe in everything. Or he believes in one thing at a time. He believes a second thing only if it somehow follows from the first thing. He is nearsighted and methodical, avoiding wide horizons. If two things don't fit, but you believe both of them, thinking that somewhere, hidden, there must be a third thing that connects them, that's credulity.And personally, that is the goal of my intellectual efforts.
Incredulity doesn't kill curiosity, it encourages it. (pp. 49-50)