For more than fifty years, the Solow decomposition (Solow 1957) has served as the standard measurement of total factor productivity (TFP) growth in economics and management, yet little is known about its precision, especially when the capital stock is poorly measured. Using synthetic data generated from a prototypical stochastic growth model, we explore the quantitative extent of capital measurement error when the initial condition is unknown to the analyst and when capacity utilization and depreciation are endogenous. We propose two alternative measurements which eliminate capital stocks from the decomposition and significantly outperform the conventional Solow residual, reducing the root mean squared error in simulated data by as much as two-thirds. This improvement is inversely related to the sample size as well as proximity to the steady state. As an application, we compute and compare TFP growth estimates using data from the new and old German federal states.