In most well-researched topics, there are disparate, often contradictory, empirical findings. A meta-analysis is a quantitative literature review that, according to Pedro Neves (MACGROW), applies “objective techniques to find patterns in those divergences and explain the factors behind them.” Pedro has conducted several, especially in his field of expertise, economic growth. Here, “as in other areas, there are a lot of disparate results in the empirical literature”.
As Pedro highlights, a typical meta-analytical study of a particular effect has three stages. First, after gathering all the relevant literature, a weighted average effect is calculated to get a new estimate of sign and magnitude. Second, if needed, publication bias, resulting from the fact that studies with significant effects are more likely to get published, is accounted for. “In many cases there are very substantial biases,” Pedro warns, “and the initial average effect ceases to be significant.” Finally, through a regression, the sources of heterogeneity among empirical results are investigated, using independent variables like the type of data used in the primary study, estimation method, or time horizon.
An example of Pedro’s meta-analytical works is an article written with Tiago Sequeira (University of Coimbra) on spillover effects of knowledge on its production. “Endogenous growth theories view R&D as crucial for sustained economic growth,” notes Pedro, “and raise the question of the magnitude of the effect of the current stock of knowledge on the production of new knowledge.” First, the authors find that the average spillover effect is statistically significant and less than one (around 0.8). Second, they find no evidence of publication bias. In Pedro’s words, these findings “lend some support to the semi-endogenous growth models”, while also “providing an objective and reliable value for parameter calibration.” Finally, the authors find that estimates tend to be lower in studies focusing on rich economies, or only using regional data, whereas they tend to be higher when estimation accounts for foreign influence on knowledge production or when industrial data are used. Thus, the heterogeneity across studies suggests the “need for different R&D policies across sectors or industries.”
Besides his interest in meta-analytical work, Pedro is currently exploring several topics, like the interplay between poverty and economic growth, or the underlying drivers of youth unemployment.
