change directory structure and have final files as "index.qmd"

crress/_quarto.yml

@@ -30,23 +30,23 @@ book:
     - about.qmd
     - part: sessions/session1/session1_intro.qmd
       chapters:
-        - sessions/session1/converted/salmon/salmon.qmd
-        - sessions/session1/converted/whited/whited.qmd
+        - sessions/session1/salmon/index.qmd
+        - sessions/session1/whited/index.qmd
     - part: sessions/session2/session2_intro.qmd
       chapters:
-        - sessions/session2/converted/swauger/swauger.qmd
+        - sessions/session2/swauger/index.qmd
     - part: sessions/session3/session3_intro.qmd
       chapters:
-        - sessions/session3/converted/diego/diego.qmd
+        - sessions/session3/diego/index.qmd
     - part: sessions/session4/session4_intro.qmd
       chapters:
-        - sessions/session4/converted/guimaraes/guimaraes.qmd
+        - sessions/session4/guimaraes/index.qmd
     - part: sessions/session5/session5_intro.qmd
       chapters:
-        - sessions/session5/converted/hoynes/hoynes.qmd
+        - sessions/session5/hoynes/index.qmd
     - part: sessions/session7/session7_intro.qmd
       chapters:
-        - sessions/session7/converted/macdonald/macdonald.qmd
+        - sessions/session7/macdonald/index.qmd
 format:
   html:
     theme: 

crress/sessions/session1/whited/index.qmd

@@ -0,0 +1,126 @@
+---
+author:
+- affiliations: Ross School of Business at the University of Michigan and Editor of
+    Journal of Financial Economics
+  name:
+  - Toni M. Whited
+title: Comments on Reproducibility in Finance and Economics
+---
+
+## Introduction
+
+Reproducibility is defined as obtaining consistent results using the
+same data and code as the original study. Most of the discussion of
+reproducibility has centered around the many obvious benefits.
+Reproducible research advances knowledge for several reasons. It reduces
+the risk of errors. It also makes the processes that generate results
+more transparent. This second advantage has an important educational
+component, as it helps disseminate not just results but processes.
+However, reproducibility is not without costs. Good research procedures
+consume resources both in terms of a researcher's own efforts and in
+terms of the involvement of arms-length parties in actually reproducing
+the research. This second cost is not just a time cost; it is pecuniary
+as well.
+
+Thus, reproducibility is a good that is costly to produce and that has
+many positive externalities. Researchers internalize many of the
+benefits of reproducibility, especially in terms of research
+extendability and personal reputation. However, they do not internalize
+any of the benefits to the research community at large. Because
+reproducibility is costly, it is unlikely to be produced at a socially
+optimal rate by any individual researchers. Thus, the questions are the
+extent to which reproducibility should be subsidized and who should
+subsidize it. Should all research be reproduced by arms-length parties,
+and what are the least costly policies that facilitate reproducible
+research? The rest of this note is organized around policies regarding
+actual reproduction and proprietary data.
+
+## Code, Data, and Arms-Length Reproduction
+
+One low-cost and easily implementable set of policies that enhances the
+reproducibility of research is journals' data and code disclosure
+policies. In the age of inexpensive data storage and an abundance of
+public repositories, the costs of these policies are small, and the
+policies should be implemented. They impose some costs on researchers in
+terms of organizing data and code, but well-organized data and code are
+already an essential part of the research process, so these costs should
+be small.
+
+While simple to implement, this low-cost policy is not without
+non-pecuniary drawbacks for journals. The code and data can be
+incomplete, poorly documented, or unusable. Moreover, journal editors
+have to retract articles that, after publication, cannot be reproduced.
+In economics, these concerns have prompted journals to start arms-length
+reproduction of results. The benefit of this policy is primarily that
+authors and journals can be confident that the code submitted with an
+article actually works to reproduce the results.
+
+However, the pecuniary costs of this policy can be substantial. It is
+expensive for journals to hire data editors and well-trained research
+assistants, and many academic journals run on tight budgets. It is often
+time-consuming for authors to comply with reproducibility requirements.
+This last issue is particularly burdensome for authors who cannot afford
+research assistance.
+
+While the above issues involve costs, the following are more
+fundamental. Reproducibility policies give researchers incentives to do
+research that is easier to reproduce, thus restraining research
+innovation that requires either large data or intense computing. Most
+importantly, code that can run on data and reproduce results can still
+contain errors.
+
+These arguments imply that while individual researchers are likely to
+underproduce reproducibility, it is also unlikely optimal for the
+progress of science that all research be reproduced before publication.
+Some papers, even those in the very best journals, rarely get read or
+cited, and the benefits of reproducing these papers are small.
+
+However, ex-ante, it is hard to know which papers will attract attention
+and which will not. One solution that lies between data and code
+disclosure and arms-length reproduction is verification. It is much less
+expensive to verify the contents of a replication package than to do an
+actual reproduction. Verification might consist of checking for the
+existence of replication instructions, an execution script, or either
+data or pseudo-data. This type of service could be provided by journals
+or other third parties, much as copy editors fix syntax and grammar
+errors before articles are submitted. At that point, reproducibility
+would be left up to the academic community, with the more important
+pieces of research being subject to greater scrutiny.
+
+A final issue with reproducibility is education. In economics and
+finance, students are not taught how to create reproducible research. An
+improvement that would go a long way toward improving the culture
+surrounding reproducibility would be to teach PhD students how to
+organize research projects and to write code in such a way that others
+can reproduce results easily. This type of education would lower the
+costs to individual researchers of making their own research
+reproducible.
+
+## Proprietary Data
+
+A possibly larger challenge for reproducibility than verification or
+arms-length execution of code is proprietary data. A clarification is
+necessary because not all types of data with restricted access are
+completely secret, that is, available only to the data provider and a
+researcher. For example, commercial data sets are not secret, just
+costly to obtain. Similarly, administrative datasets are not secret.
+They just require special permission. In contrast, proprietary data
+cannot be offered to the research community at large for the purposes of
+reproducing the results. So the question is whether journals should
+discourage the use of this type of data or require that verifiers have
+access to the data. Given the large number of studies using proprietary
+data, this issue is possibly more important than the issue of running
+code.
+
+## Conclusion
+
+In conclusion, the reproducibility of research is essential for the
+advancement of science. However, it is not without costs, so blanket
+statements that all research should be reproducible are not feasible.
+Instead, feasible policies include those that lower the costs for others
+to replicate research. Data and code disclosure is a low-cost policy
+that should be implemented widely. Verification of code and data
+packages is a slightly more costly option. Arms-length reproduction is a
+much more costly alternative. Finally, perhaps the most important issue
+that impedes reproducibility in finance and economics is the use of
+proprietary data.