What determines the distribution of establishments in terms of size and life-cycle growth? How are those determinants related to aggregate productivity? We provide novel answers by developing a framework that uses price and quantity information on establishments' outputs and inputs to jointly estimate the demand and production parameters, and subsequently establishments' quality-adjusted productivity, deriving both micro-level and aggregate implications. We find that the dominant source of variation in establishment size is variation in quality/product appeal but that variation in technical efficiency plays an important supporting role. Multiple factors dampen dispersion in establishment size including dispersion in input (quality-adjusted) prices, markups, and residual wedges. Relatively moderate dampening factors induce large aggregate allocative efficiency losses relative to their absence. We show that joint estimation of the parameters of the demand and production function crucially affects inferences on the determinants of the size distribution of firms and their implications for aggregate productivity.
Using official employment surveys for 45 advanced economies and Latin American countries, this paper shows that the positive cross-country correlation between business size and GDP per capita is tighter than previously found using firm-level datasets and finds a close negative business size-Gini relationship. The paper also finds a closer connection between individual income and business size for workers in less developed countries compared with those in advanced economies. Because employment data address the bias against the smallest productive units that characterize firm-level datasets, our approach uniquely assesses and highlights the dominance of the left tail of the business size distribution in less developed countries.
Market Concentration, Market Fragmentation, and
Inequality in Latin America (2021). UNDP LAC Working Paper No. 11.
Background Paper for the UNDP LAC Regional Human Development Report
2021. With Marcela Eslava and Marcela Meléndez. [Working Paper].
Inequality in Latin America is much higher than in
Europe and the US. The income distribution in the region is also much
more skewed, displaying a thicker left tail. And, like those comparison
economies, it also exhibits a long right tail. We illustrate the link
between this bipolar character of inequality in the region and the
similarly bipolar character of the distribution of productive units,
where income is generated. The firm size distribution in Latin America
is dominated by a plethora of tiny businesses, which absorb several
times more employment than in the US and Europe and exhibit a much lower
relative productivity, while its upper tail exhibits higher market
concentration. 34% of the distance in the 50/10 personal income gap
between Latin America and the US is explained by a higher concentration
of workers in the categories with poorer relative income and
productivity: self-employment and employment in micro establishments. In
the right tail of the income distribution, the larger 90/50 personal
income gap in the region compared to the US is fully explained by the
larger relative income of business owners of large firms with high
relative markups. We show that market concentration in the region is
large and tied to small labor shares. It is also closely tied to the
extreme dispersion of productivity and the prevalence of low
productivity businesses. The central message is that high inequality in
the region is deeply rooted in the productivity problem.
"Talented" Men and "Kind" Women? Gender Bias in Student Evaluations of Teaching (2021). Documento CEDE No. 46. (Master's Thesis) [Working
Paper]. Media coverage: [El Espectador].
This paper studies gender biases in the student
evaluations of teaching at a university in Colombia. I use a linear
regression with student fixed effects that account for grades and
professor skills, measured with a value-added model, to determine that,
on average, full-time female professors receive evaluations lower by
0.07 standard deviations than male professors. Meanwhile, there are no
gender differences among adjunct professors or teacher assistants.
However, there is significant heterogeneity. In courses with 35 to 55
year-old professors, with more than 50 students, or with a smaller share
of female students, female professors are discriminated against the
most. In such cases, biases have a magnitude of about 0.1-0.2 standard
deviations and represent 40\% of all evaluations. Additionally, women
are mostly described with words related to personal characteristics
("kind" or "comprehensive"), and men are associated with course-related
topics or positive adjectives ("topic", "exam", "talented"). Based on
the results found, I propose several possible measures to reduce student
bias and allow analysis on the evaluations to consider the gender
differences.