R not detecting missing values

[1]In my dataset I have responses to a yes/no question, with a lot of missing values.

The column for the question looks something like this:

Question[1] yes[2] no[3] [4] yes[5] no[6] 

In other words:

summary(Question)173yes160no155

where we have 173 missing values, 160 yes answers, and 155 no's.

When I look at levels in the factor, I get the following:

levels(Question)[1] " "[2] yes[3] no

I would like to drop the missing values (that is, level " ") (and have legitimate reasons to exclude missing values in this case).

However, is.na(Question) reports (implausibly) that there are no missing values, so I cannot easily exclude them.

I have tried dropping the level with missing values:

droplevels.factor(Question, exclude={" "}

but it results in a "NAs introduced by coercion" warning message.

What can I do to exclude the level with missing values? Please help. Thank you.

Edited with link to data file.