Evidence of Systematic Errors in $Spitzer$ Microlens Parallax Measurements

The microlensing parallax campaign with the $Spitzer$ space telescope aims to measure masses and distances of microlensing events seen towards the Galactic bulge, with a focus on planetary microlensing events. The hope is to measure how the distribution of planets depends on position within the Galaxy. In this paper, we compare 50 microlens parallax measurements from the 2015 $Spitzer$ campaign to three different Galactic models commonly used in microlensing analyses, and we find that $\geq 74\%$ of these events have microlensing parallax values higher than the medians predicted by Galactic models. The Anderson-Darling tests indicate probabilities of $p_{\rm AD} < 6.6 \times 10^{-5}$ for these three Galactic models, while the binomial probability of such a large fraction of large microlensing parallax values is $< 4.6\times 10^{-4}$. Given that many $Spitzer$ light curves show evidence of large correlated errors, we conclude that this discrepancy is probably due to systematic errors in the $Spitzer$ photometry. We find formally acceptable probabilities of $p_{\rm AD} > 0.05$ for subsamples of events with bright source stars ($I_{\rm S} \leq 17.75$) or $Spitzer$ coverage of the light curve peak. This indicates that the systematic errors have a more serious influence on faint events, especially when the light curve peak is not covered by $Spitzer$. We find that multiplying an error bar renormalization factor of 2.2 by the reported error bars on the $Spitzer$ microlensing parallax measurements provides reasonable agreement with all three Galactic models. However, corrections to the uncertainties in the $Spitzer$ photometry itself are a more effective way to address the systematic errors.