Measurement-based load model parameter estimation uses measurements from a disturbance on the grid. Those measurements can include voltage and/or real and reactive power. In this paper, we show that the type of measurements used directly impacts the accuracy of parameter estimation. We look at four scenarios. With wide-area deployment of voltage sensors, such as PMUs, the resulting parameter estimation is very accurate at high signal-to-noise ratios (SNR), but is very poor at low SNRs, because voltage has low sensitivity to the parameters. With only local deployment of complex power sensors, the estimate is worse than the first scenario at all SNRs. However, with wide-area deployment of complex power sensors, the estimate becomes very robust to low SNR, because complex power has much higher sensitivity to the parameters. Combining wide-area voltage and power measurements produces the best results.