One problem confronting the use of citation-based metrics in science studies and research evaluations is the Matthew effect. This paper reviews the role of citations in science and decomposes the Matthew effect in citations into three components: networking, prestige, and appropriateness. The networking and prestige effects challenge the validity of citation-based metrics, but the appropriateness effect does not. Using panel data of 1279 solo-authored papers' citation histories and fixed effects models, we test these three effects controlling for unobserved paper characteristics. We find no evidence of retroactive networking effect and only weak evidence of prestige effect (very small and not always significant), which provides some support for the use of citation-based metrics in science studies and evaluation practices. In addition, adding the appropriateness effect reduces the size of the prestige effect considerably, suggesting that previous studies controlling for paper quality but not appropriateness may have overestimated the prestige effect.