“…In either impact model, we include journal characteristics that could be correlated with the journal impact factor: (i) the fields of research of the journal, given by the shares of articles published in each of the 19 JEL codes 32 ; (ii) the number
of articles published, which captures the capacity of the journal and could be negative related to journal quality (Ductor et al.,
2020); (iii) average article length,
, as it is positively related to citations (Bramoullé & Ductor,
2018; Card & DellaVigna,
2013); and (iv) the average number of authors,
, of an article. In the panel regression, we use yearly values of these variables, marked with a subscript
, while in the cross‐sectional analysis, we use the average yearly values, marked with an overline.…”