The transverse proximity effect is the expected decrease in the strength of the Lya forest absorption in a QSO spectrum when another QSO lying close to the line of sight enhances the photoionization rate above that due to the average cosmic ionizing background. We select three QSOs from the Early Data Release of the Sloan Digital Sky Survey that have nearby foreground QSOs, with proper line of sight tangential separations of 0.50, 0.82, and 1.10 h^{-1} Mpc. We estimate that the ionizing flux from the foreground QSO should increase the photoionization rate by a factor (94, 13, 13) in these three cases, which would be clearly detectable in the first QSO and marginally so in the other two. We do not detect the transverse proximity effect. Three possible explanations are provided: an increase of the gas density in the vicinity of QSOs, time variability, and anisotropy of the QSO emission. We find that the increase of gas density near QSOs can be important if they are located in the most massive halos present at high redshift, but is not enough to fully explain the absence of the transverse proximity effect. Anisotropy requires an unrealistically small opening angle of the QSO emission. Variability demands that the luminosity of the QSO with the largest predicted effect was much lower 10^6 years ago, whereas the transverse proximity effect observed in the HeII Lya absorption in QSO 0302-003 by Jakobsen et al. (2003) implies a lifetime longer than 10^7 years. A combination of all three effects may better explain the lack of Lya absorption reduction. A larger sample of QSO pairs may be used to diagnose the environment, anisotropy and lifetime distribution of QSOs.

We determine the evolution of the faint, high-redshift, optical luminosity function (LF) of AGN implied by several observationally-motivated models of the ionizing background. Our results depend crucially on whether we use the total ionizing rate measured by the proximity effect technique or the lower determination from the flux decrement distribution of Ly alpha forest lines. Assuming a faint-end LF slope of 1.58 and the SDSS estimates of the bright-end slope and normalization, we find that the LF must break at M_B*=-24.2,-22.3, -20.8 at z=3,4, 5 if we adopt the lower ionization rate and assume no stellar contribution to the background. The break must occur at M_B*=-20.6,-18.7, -18.7 for the proximity effect background estimate. These values brighten by as much as ~2 mag if high-z galaxies contribute to the background with an escape fraction of ionizing photons consistent with recent estimates: f_e=0.16. By comparing to faint AGN searches, we find that the typically-quoted proximity effect estimates of the background imply an over-abundance of faint AGN (even with f_e=1). Even adopting the lower bound on proximity effect measurements, the stellar escape fraction must be high: f_e>0.2. Conversely, the lower flux- decrement-derived background requires a limited stellar contribution: f_e<0.05. Our derived LFs together with the locally-estimated black hole density suggest that the efficiency of converting mass to light in optically-unobscured AGN is somewhat lower than expected, <0.05. Comparison with similar estimates based on X-ray counts suggests that more than half of all AGN are obscured in the UV/optical. We also derive lower limits on typical AGN lifetimes and obtain >10^7 yrs for favored cases.