Methods

  • Big Data in economics

    New sources of data create challenges that may require new skills

    Big Data refers to data sets of much larger size, higher frequency, and often more personalized information. Examples include data collected by smart sensors in homes or aggregation of tweets on Twitter. In small data sets, traditional econometric methods tend to outperform more complex techniques. In large data sets, however, machine learning methods shine. New analytic approaches are needed to make the most of Big Data in economics. Researchers and policymakers should thus pay close attention to recent developments in machine learning techniques if they want to fully take advantage of these new sources of Big Data.
    MoreLess
  • Measuring employment and unemployment

    Should statistical criteria for measuring employment and unemployment be re-examined?

    Measuring employment and unemployment is essential for economic policy. Internationally agreed measures (e.g. headcount employment and unemployment rates based on standard definitions) enhance comparability across time and space, but changes in real labor markets and policy agendas challenge these traditional conventions. Boundaries between different labor market states are blurred, complicating identification. Individual experiences in each state may vary considerably, highlighting the importance of how each employed or unemployed person is weighted in statistical indices.
    MoreLess
  • The usefulness of experiments

    Are experiments the gold standard or just over-hyped?

    Jeffrey A. Smith, May 2018
    Non-experimental evaluations of programs compare individuals who choose to participate in a program to individuals who do not. Such comparisons run the risk of conflating non-random selection into the program with its causal effects. By randomly assigning individuals to participate in the program or not, experimental evaluations remove the potential for non-random selection to bias comparisons of participants and non-participants. In so doing, they provide compelling causal evidence of program effects. At the same time, experiments are not a panacea, and require careful design and interpretation.
    MoreLess
  • Relative deprivation in the labor market

    The choice of reference group crucially determines subjective deprivation and thus affects labor market behavior

    Paolo Verme, June 2017
    Why do different population groups (e.g. rural vs. urban, youth vs. elderly and men vs. women) experience the same objective labor status differently? One hypothesis is that people are more concerned with relative deprivation than objective deprivation and they value their own status relative to the status of their peers—the reference group. One way to test this hypothesis in the labor market is to measure individual differences in labor status while controlling for characteristics that define population groups. This measure is called “relative labor deprivation” and can help policymakers to better understand how labor claims are generated.
    MoreLess
  • Gross domestic product: Are other measures needed?

    GDP summarizes only one aspect of a country’s condition; other measures in addition to GDP would be valuable

    Gross domestic product (GDP) is the key indicator of the health of an economy and can be easily compared across countries. But it has limitations. GDP tells what is going on today, but does not inform about sustainability of growth. It does not measure happiness, so residents can be dissatisfied even when GDP is rising. GDP does not consider environmental factors or reflect what individuals do outside paid employment. It might increase in times of military conflicts and after natural disasters or terrorist acts, as the loss of property is not counted. Hence, complementary measures may help to show a more comprehensive picture of an economy.
    MoreLess
  • Identifying and measuring economic discrimination

    Using decomposition methods helps measure both the amount and source of economic discrimination between groups

    Sergio Pinheiro Firpo, March 2017
    Differences in wages between men and women, white and black workers, or any two distinct groups are a controversial feature of the labor market, raising concern about discrimination by employers. Decomposition methods shed light on those differences by separating them into: (i) composition effects, which are explained by differences in the distribution of observable variables, e.g. education level; and (ii) structural effects, which are explained by differences in the returns to observable and unobservable variables. Often, a significant structural effect, such as different returns to education, can be indicative of discrimination.
    MoreLess
  • Using linear regression to establish empirical relationships

    Linear regression is a powerful tool for estimating the relationship between one variable and a set of other variables

    Marno Verbeek, February 2017
    Linear regression is a powerful tool for investigating the relationships between multiple variables by relating one variable to a set of variables. It can identify the effect of one variable while adjusting for other observable differences. For example, it can analyze how wages relate to gender, after controlling for differences in background characteristics such as education and experience. A linear regression model is typically estimated by ordinary least squares, which minimizes the differences between the observed sample values and the fitted values from the model. Multiple tools are available to evaluate the model.
    MoreLess
  • Maximum likelihood and economic modeling

    Maximum likelihood is a general and flexible method to estimate the parameters of models in labor economics

    Gauthier Lanot, January 2017
    Most of the data available to economists is observational rather than the outcome of natural or quasi experiments. This complicates analysis because it is common for observationally distinct individuals to exhibit similar responses to a given environment and for observationally identical individuals to respond differently to similar incentives. In such situations, using maximum likelihood methods to fit an economic model can provide a general approach to describing the observed data, whatever its nature. The predictions obtained from a fitted model provide crucial information about the distributional outcomes of economic policies.
    MoreLess
  • Can lab experiments help design personnel policies?

    Employers can use laboratory experiments to structure payment policies and incentive schemes

    Marie Claire Villeval, November 2016
    Can a company attract a different type of employee by changing its compensation scheme? Is it sufficient to pay more to increase employees’ motivation? Should a firm provide evaluation feedback to employees based on their absolute or their relative performance? Laboratory experiments can help address these questions by identifying the causal impact of variations in personnel policy on employees’ productivity and mobility. Although they are collected in an artificial environment, the qualitative external validity of findings from the lab is now well recognized.
    MoreLess
  • Meta-regression analysis: Producing credible estimates from diverse evidence

    Meta-regression methods can be used to develop evidence-based policies when the evidence base lacks credibility

    Chris Doucouliagos, November 2016
    Good policy requires reliable scientific knowledge, but there are many obstacles. Most econometric estimates lack adequate statistical power; some estimates cannot be replicated; publication selection bias (the selective reporting of results) is common; and there is wide variation in the evidence base on most policy issues. Meta-regression analysis offers a way to increase statistical power, correct the evidence base for a range of biases, and make sense of the unceasing flow of contradictory econometric estimates. It enables policymakers to develop evidence-based policies even when the initial evidence base lacks credibility.
    MoreLess
show more