Can central counterparty (CCP) clearing control counterparty risk in the presence of risk taking that can aggravate such risk? When counterparty risk is not observable, I show that central clearing leads to higher collateral requirements for two different reasons. Without collusion about risk taking, a CCP offering diversication of risk cannot selectively forgo incentives for transactions that use collateral only for insurance. With collusion about risk taking, a CCP needs to charge collateral in line with the worst counterparty quality to control risk taking. Requiring more collateral reduces market liquidity and worsens incentives causing a feedback effect that amplifies collateral costs.