Many people derive a sense of impact or purpose from their jobs – they consider work to be a source of meaning. But what makes work meaningful? Theoretical models suggest that meaning can be created through social and non-social impact. We exploit rich panel data representative of workers in the United States to empirically assess these models, and estimate a non-linear production function for work meaning that allows for noisy and complementary inputs. The model explains most of the variation in meaning between and half of the variation in meaning within individuals. We find that social impact is the most effective pathway to meaning, and estimate a direct output elasticity with respect to work meaning of about 0.44. The effectiveness of non-social impact differs significantly across demographic groups. We also find evidence of a negative interaction between social- and non-social impact. A standard deviation increase in social impact is twice as effective in creating meaning for individuals that perceive their jobs as having little non-social impact, compared to those with high perceived non-social impact.

We introduce a teacher time allocation model in which teachers allocate their available instructional time among individual, group, and classroom instruction to maximize welfare function of all students’ test score. Teachers allocate time based on their perceptions of pupil productivity. We consider two variants of the model: one with knowledge spillovers and one with instruction spillovers. We conduct a survey among primary school teachers in Flanders, asking for each pupil’s productivity, test score, and time allocation in mathematics. We use the data to evaluate both variants and find that the model with instruction spillovers fits the observed behavior of the teachers better but requires more assumptions. We also derive teachers’ marginal social welfare weights for their pupils and examine the factors influencing them. The weights are predominantly positive, indicating teacher efficiency, decrease with higher math scores, suggesting inequality aversion, and show no significant correlation with gender, home language, or mother’s education, implying anonymity. These results appear robust regardless of the presence and type of spillover effects.