In this talk, I will present our recent works studying the effects of centralization and transparency across multiple layers of blockchain-based systems. Centralization in blockchains can manifest differently across consensus, protocol, and application layers, challenging the simplistic view of decentralized versus centralized systems. I will start by presenting our mathematical model of centralization effects in proof-of-work mining, which quantifies how hash rate concentration and block propagation dynamics affect fork rates and resource efficiency in distributed ledgers. I will also discuss our method for the statistical detection of selfish mining in several proof-of-work blockchains, which offered the first empirical evidence of coordinated strategic behaviour in mining. I will then move to the application layer to present our study on DEX liquidity dynamics during the Silicon Valley Bank collapse, where we focus on the counterintuitive effects of transparency during market stress. We show how USDC’s frequent disclosures led to swift market reactions and reduced dominance, while USDT’s opacity provided relative stability through delayed information propagation. Our empirical analysis builds on a successful stream of financial economics literature about how transparency regimes affect market dynamics. Our work encourages a more critical assessment of the economics of blockchain-based systems. I will conclude by discussing my views on where we should look in the coming years to design appropriate risk management practices for users, developers and regulators, particularly in the context of the growing DeFi-TradFi integration.