Media outlets are increasingly owned by conglomerates, inducing a conflict of interest: a media outlet can bias its coverage to benefit companies in the same group. We test for bias by examining movie reviews in media outlets owned by News Corp, such as the Wall Street Journal, and Time Warner, such as Time. We find higher ratings for 20th Century Fox movies in News Corp. outlets compared to movies by other studios. To disentangle bias from correlation of taste, we introduce and validate a novel matching procedure using individual movie ratings from online platforms. Using this procedure, we find no evidence of bias in News Corp. nor Time Warner outlets. We reject even small effects, such as bias of one extra star (out of four) every thirteen movies. We test for differential bias when the return to bias is plausibly higher, examine bias by media outlet and by journalist, as well as editorial bias. We also consider bias by omission—whether media outlets are more likely to review highly-rated movies by affiliated studios—and conflict of interest within a movie aggregator. In none of these dimensions do we find evidence of bias. We relate to previous work and discuss three explanations for the lack of bias in our setting: high values of media reputation, organizational features in a conglomerate, and low returns to bias.

We propose a model of job search with reference-dependent preferences, with loss aversion relative to recent income (the reference point). In this model, newly unemployed individuals search hard since consumption is below their reference point. Over time, though, they get used to lower income and thus reduce their search effort. In anticipation of a benefit cut, their search effort rises again, then declines once they get accustomed to the lower postcut benefit level. The model fits the typical pattern of exit from unemployment, even with no unobserved heterogeneity. To distinguish between this and other models, we use a unique reform in the unemployment insurance (UI) benefit path. In 2005, Hungary switched from a single-step UI system to a two-step system, with overall generosity unchanged. The system generated increased hazard rates in anticipation of, and especially following, benefit cuts in ways the standard model has a hard time explaining. We estimate a model with optimal consumption, endogenous search effort, and unobserved heterogeneity. The reference-dependent model fits the hazard rates substantially better than plausible versions of the standard model, including habit formation. Our estimates indicate a slow-adjusting reference point and substantial impatience, likely reflecting present-bias.