We are doing research that could not be done even a decade ago. It wasn't possible.

From Sex, Ancestral, and pattern type variation of fingerprint minutiae: A forensic perspective on anthropological dermatoglyphics by Nichole A. Fournier and Ann H. Ross.  From the Abstract:

The effect of sex is insignificant. Ancestry is significant (Wilks' λ = 0.053 F value = 2.98, DF = 4,224, P value = 0.02), as is pattern type (Wilks' λ = 0.874 F value = 2.57, DF = 12,592.94, P value = 0.003). The ANOVA reveals that bifurcations are responsible for the variation between ancestral groups, while bifurcations and ending ridges vary between patterns. Logistic regression results suggest that total bifurcations can predict the ancestry of an individual (ChiSq = 6.55, df = 1, Prob > ChiSq = 0.01).

Translating from Academic, I believe they are saying that relying on fingerprints, you cannot identify the individual's sex but you can identify their racial/ethnic ancestry.

Whether it is a well founded conclusion or not, I have no opinion other than to note a pitiably small sample size.  

It is just striking to me that such research could only have occurred in the past decade or so.  Prior to that, we did not have the computing power or machine learning capacity to conduct such research.