Correlation

#1
Hi all,
I am running a test using a binary variable representing the presence (or not) of a theoretical construct. The other variable I have is the age of the respondent (5 - 19 years old).
If I treated age as a continuous variable, I would use point-biserial correlation. However, the theory I am using uses four age groups (3-5, 6-8, 9-13, and 14+) making it ordinal. Is rpb the best choice for binary/ordinal data as well? If not, what is the better choice?
OR, I can use age as a continuous variable, but this would threaten the validity of my paper if I don't stick to the theory unless I can justify it (such as no reliable test between binary and ordinal).
Thanks!