If you only had five minutes: Best advice for new instructors of economics

Abstract

Teaching a course in economics for the first time can be a daunting task, whether the instructor is a graduate student or a new faculty member in their first post-PhD years. In gauging what advice is considered most vital from among the plethora of potential sources, the authors surveyed seasoned economics instructors, asking respondents to distill their advice into what they could provide within a five-minute time constraint. Their responses were then processed with a mix of human perception and machine-based natural language processing. In this manner, they gained an understanding of what seems to matter the most when starting out in the economics classroom and thus provide usable insights for economic educators on both the giving and receiving end of the guidance process.

Keywords:

• Advice
• new instructors
• teaching economics

JEL CODES:

• A20
• A22

Acknowledgments

The authors thank Chase Coleman for excellent research assistance, Jose Fernandez for invaluable feedback, and participants in the 2022 ASSA conference session “If You Only Had Two Hours—Best Advice for New Instructors of Economics” for helpful comments.

Notes

1 Our survey received IRB approval from Smith College, Protocol #21-004. Appendix 1 shows the complete survey instrument.

2 As it turned out, the survey respondents were evenly spread between colleges and universities. While some differences did emerge on how to implement some suggestions due to class size, no particular distinctions across institution type in terms of key themes were evident.

3 https://esrc.ukri.org/about-us/what-is-social-science/qualitative-research/

4 We received permission from these survey participants to quote their responses with attribution.

5 Specifically, each survey response was converted into a single string and stripped of punctuation and nonalphanumeric characters. Individual words were stemmed and lemmatized to their base form to account for different forms or conjugations of a word. (For example, “econom” captures “economic” or “economics” or “economist” while “think” covers “think,” “thinks,” “thinkers,” or “thinking.”) Stop words, as defined by the Natural Language Toolkit Package (see Bird, Loper, and Klein 2009), were removed. These are extremely common words, such as “the,” “a,” “on,” and “of,” that don’t carry any meaning in the analysis.

6 Latent Dirichlet Allocation (LDA) is another natural language processing technique commonly used to identify the most frequently occurring words or groups of words in responses. It can reduce a large corpus of text into distinct topics. Such unsupervised machine-learning-based statistical models are typically used for analyzing large, unstructured collections of documents. Given the small size of our data sample, LDA is not a suitable tool in this case.

7 The topmost frequently used words are broadly the same in subsamples based on gender (male [57%], female [41%]) and institution type (research universities [57%], colleges [43%])—the one exception being that time, rather than course, makes it to the top five in the research institutions ranking. The percentages of the full sample represented by these categories are in brackets.

8 Figures A2.1–A2.8 in appendix 2 show the unweighted and weighted bigram rankings for male and female respondents and for respondents at colleges versus research universities. The gender breakdown between male and female is based on 98 of the 100 responses.

9 It is interesting to note the correspondence between these findings for LAC respondents and the findings for the female subsample. While at first glance, one might think this pattern arises because of a significant overlap in the composition of the LAC and female subgroups, this is not the case. The 41 female respondents were approximately equally spread across LAC and research universities.