Instrumental variables are the class of variables that only have direct effects on the explanatory variables and are uncorrelated with the error term. So that they can be used to isolate the exogenous portion of movements of explanatory variables in a regression. This characteristic makes it a powerful tool for econometricians to cope with the endogeneity problems and identify the causal relationship between key economic factors from primitive statistical correlations. Therefore, since being introduced in the 1920s, it has been widely used by econometricians and become a common identification strategy in applied researches. This essay gives a critical assessment on the application of this approach in the paper “Why Do Some Countries Produce So Much More Output per Worker than Others” (Hall & Jones, 1999).
As indicated in the title, the object of the Hall and Jones  paper is to explain the vast differences in economic performance—captured by output per worker—among countries around the world. This question puzzled generations of economists in the past several decades and created a vibrant literature in applied research of developmental economics. Conventionally, economists decomposed the output per worker into three ingredients: the stock of physical capital, the human capital-augmented labor, and the productivity which is usually captured by Solow residuals. Hence the enormous difference in output per worker should be explained by these three factors. Although Hall and Jones  also followed this growth accounting approach, they argue that this approach only captured the immediate cause of economic development.
From the discuss above we can see that, although it is a novel idea to construct instruments based on the course of history, it is a tricky and never-ending task to accurately capture this historical force. There are many detailed events and policies that need to be taken into consideration. In regard to the instrumental variable estimation as an identification strategy, its power comes from our profound knowledge of the relevant factors. It will be a mistake to adopt an instrument solely by speculation. We need to think real hard about the connection of various factors, consider the possibility of different causal chain, and carefully examine our hypothesis by resorting to empirical tests.