We derive empirical predictions from the standard investment-cash flow framework on the functioning of internal capital markets (ICM), but circumvent its criticism by focusing on parent cash flow and investment opportunities. We test these predictions using a unique data set of parent firms and their listed and unlisted subsidiaries in 90 countries over the period 1995-2006. We find that company and country institutional structures matter. (1) Ownership participation of the parent firm in the subsidiary plays a crucial role for the proper functioning of ICMs. The larger the ownership stake of the parent, the better the functioning of the ICM. (2) The best functioning cross-border ICMs can be found in the sub-sample of firms with parents from a country with "strong" institutions and subsidiaries from a country with "weak" institutions. (3) Unlisted subsidiaries are much more dependent on the ICMs their parents provide than listed subsidiaries. Thus, ICMs are not per se "bright" or "dark", their proper functioning depends on how they are set up. (authors' abstract)