Abstract: Theories of microfoundations for aggregate volatility rely on a skewed in- dividual size distribution, termed granularity (in the sense that there are big firms that can’t be subdivided). If so, what causes granularity? The firm size distribution may be skewed because of a skewed productivity distribution, geographic agglomeration, or skewed demand characteristics that result in an asymmetric production network. What matters more? I use detailed data on firm-firm trade in Canada to estimate a model in which productivity, geography, and demand characteristics vary independently. This allows me to recover un- observed demand characteristics from the observed production network, which conflates productivity, demand and geography. I find that the demand network accounts for 60% of the firm size distribution, productivity and geography ex- plain little, and that approximately half of the demand network effect is due to higher order network interconnections. Microeconomic shocks can account for approximately 32% of aggregate volatility, and removing variation in the demand network would reduce aggregate volatility by 25%.
Abstract: In the presence of often-cited provincial non-tariff trade barriers, one should observe provincial border effects in Canada. However, using provincial trade data leads to upward biased estimates of the border effect, because intra-provincial trade is skewed towards short distance flows that are poorly estimated by gravity models. We overcome this bias by using sub-provincial trade flows generated from a transaction-level transportation dataset. The results show that border effects fall as geographies are more fine-grained and uniform. In contrast to the U.S., where state border effects were eliminated using similar approaches, provincial border effects remain, with an implied 6.9% tariff equivalent.
Abstract: Due to its association with cross-country business cycles, propagation of idiosyncratic shocks, and even financial contagion, firm comovement is an important facet of macroeconomic research. However, we know little about whether pairs of establishments within firms comove more than pairs of establishments from different firms. Using a long panel of Canadian manufacturing establishments, I investigate the correlations and covariances of within-firm pairs of establishments and decompose them into labour inputs, intermediates and profit. I find that within-firm establishment pairs have correlations 0.0477 higher than between-firm establishment pairs (which have an average correlation very close to zero) after controlling for industry and region effects. A back-of-the-envelope calculation suggests the aggregate effect of removing those firm connections reduces aggregate sales covariance by 28%. Covarying intermediate input costs account for 49% of the within-firm comovement effect.