We present new estimates on the fraction of heavily X-ray obscured, Compton-thick (CT) active galactic nuclei (AGNs) out to a redshift of $z \leq$ 0.8. From a sample of 540 AGNs selected by mid-IR (MIR) properties in observed X-ray survey fields, we forward model the observed-to-intrinsic X-ray luminosity ratio ($R_{L_{\text{X}}}$) with a Markov chain Monte Carlo (MCMC) simulation to estimate the total fraction of CT AGNs ($f_{\text{CT}}$), many of which are missed in typical X-ray observations. We create model $N_{\text{H}}$ distributions and convert these to $R_{L_{\text{X}}}$ using a set of X-ray spectral models. We probe the posterior distribution of our models to infer the population of X-ray non-detected sources. From our simulation we estimate a CT fraction of $f_{\text{CT}}$ = $\text{0.555}^{+\text{0.037}}_{-\text{0.032}}$. We perform an X-ray stacking analysis for sources in Chandra X-ray Observatory fields and find that the expected soft (0.5-2 keV) and hard (2-7 keV) observed fluxes drawn from our model to be within 0.48 and 0.12 dex of our stacked fluxes, respectively. Our results suggests at least 50% of all MIR-selected AGNs, possibly more, are Compton-thick ($N_{\text{H}} \gtrsim$ 10$^{\text{24}}$ cm$^{-\text{2}}$), which is in excellent agreement with other recent work using independent methods. This work indicates that the total number of AGNs is higher than can be identified using X-ray observations alone, highlighting the importance of a multiwavelength approach. A high $f_{\text{CT}}$ also has implications for black hole (BH) accretion physics and supports models of BH and galaxy co-evolution that include periods of heavy obscuration.
Comment: 14 pages, 6 figures, 1 table, plus appendix figures. Accepted for publication in ApJ