Leading questions and loaded questions are very similar. Both are bad for surveys because they skew responses and thereby produce inaccurate data.
Let us imagine, for example, that you want to ask a question about tax rates in your country. An unbiased question might read something like “Would you say that income tax rates in the United States are much too high, a bit too high, just right, too low, or much too low.”
A leading question has wording that will suggest what the “right” answer is. Such a question might ask “Income tax rates in the US are about average for rich countries. Would you say that income tax rates… (and then give the same five options as above)” This clearly implies that “just right” would be the best answer.
Loaded questions are leading questions that use very emotional language. Using the example of taxes, a loaded question might read “The government takes away about 30% of your money each year. Would you say that tax rates…” This uses the emotional phrase “takes away” to push an anti-tax message and is therefore likely to get more responses that say the tax rates are too high.