There are two innovations in the paper as compared to the previous literature on democracy and growth. First, we consider not only the level of democracy, but also changes in this level in the 1970s-1990s as measured by increments of Freedom House political rights indices. Second, the distinction is made between democracy and law and order (order based on legal rules); the latter is measured by the rule of law, investors' risk and corruption indices. We discuss two interconnected threshold hypotheses: (1) in countries where law and order is strong enough, democratization stimulates economic growth, whereas in countries with poor law and order democratization undermines growth; (2) if democratization occurs under the conditions of poor law and order (so that illiberal democracy emerges), then shadow economy expands, quality of governance worsens, and macroeconomic policy becomes less prudent. We adduce a number of stylized facts to support our hypotheses. However our econometric findings are mixed: we report results that support the hypotheses as well as regressions that contradict them.