I introduce endogenous capital accumulation into an otherwise standard quantitative sovereign default model in the tradition of Eaton and Gersovitz (1981), and find that conditional on a level of debt, default incentives are U shaped in the capital stock: the economy with too small or too large amounts of capital is likely to default. In addition to an “excusable” motive for default in line with Grossman and Huyck (1989), our model also predicts an “opportunistic” motive for default in line with Kehoe and Levine (1993). The model predicts the “opportunistic” motive for default, because (1) capital is used as a consumption insurance vehicle during autarky after default, (2) installed capital within the border cannot be seized by foreign lenders, and (3) our model does not use an ad-hoc output cost of default which only penalizes default in high income states. The two different motives for default allow the calibrated model to generate defaults in “good” and “bad” times in simulation with a frequency of 38% and 62%, respectively. This is consistent with Tomz and Wright (2007)’s empirical finding that throughout history and across countries, around one third of sovereign defaults occurred in “good” times, when output is above trend, whereas most defaults occur in “bad” times. The model is calibrated to the business cycle moments of Argentina, and simulation results show that the model matches business cycle facts regarding emerging economies along other dimensions. Moreover, simulation results show that default in “good” times occurs (1) after the economy has accumulated a significantly large amount of capital and (2) when the economy faces a modestly good shock, both of which reduce the value of external borrowing but increase the value of staying in autarky. On the other hand, around defaults in “bad” times, the model economy displays typical “V” shape economic dynamics, with a collapse in absorption upon default, especially investment. Aggregating quarterly data from the model into annual frequency is found to overestimate the fraction of defaults in “good” times around twofold.