Explaining some .assign() idioms
Explaining pandas assignment, callable arguments, dictionary unpacking, and pandas 3 column expressions, with a comparison to dplyr across().
Blog
Posts and notes. From newest to oldest.
Posts and notes.
From newest to oldest.
Browse selected notes by topic.
2026
Explaining some
Processing SEC submissions data
Using the sec-submissions package to prepare Parquet data, preserve timestamp evidence, and distinguish dated audits from subsequent corrections.
Solutions for Ball and Brown (1968)
Updated, freshly executed Python Polars solutions for the Ball and Brown (1968) chapter, including the income-versus-cash-flow comparison.
Solutions for Panel data
Updated, freshly executed Python Polars solutions for the panel-data chapter, including disclosure regressions and design checks.
Getting data out of Postgres: an update
Updated benchmarks comparing PostgreSQL-to-Parquet export paths, with newer packages and a local database.
Working with date and times
Dates, times, time zones, and the difficulties of moving temporal data between R and DuckDB, using SEC filings.
Diff-in-diff as a causal design
A short comment on research design, parallel trends, and the interpretation of difference-in-differences.
The SEC’s timestamp “fix”
An update on inconsistent time zones in SEC submissions data, linking to the longer dates-and-times note.
Spirit of Goodwill
A waterfall chart of Boeing’s purchase price allocation for Spirit AeroSystems, rebuilt from the reported data.
Direct-method cash flow statements
An introduction to the June 2026 direct-method cash-flow finding, linking the existing XBRL analysis.
How common are direct-method cash flow statements?
Using SEC XBRL data and R to investigate how often companies report direct-method cash flow statements.
The R version of db2pq is on CRAN
A short introduction to the R package for exporting PostgreSQL tables to Parquet, including sources beyond WRDS.
Why include code in teaching?
Why working through code and exercises can reveal what drives a research paper's results, illustrated by the panel-data chapter.
Taxonomy of researchers
Five types of researchers, in a 2019 presentation to prospective Honours students, shared again in 2026.
What's the most important financial statement?
Using the Ball and Brown research design to examine the claim that the cash flow statement is the most important financial statement.
A mini-textbook on consolidation accounting
An introduction to freely available consolidation accounting materials, with editable Quarto sources.
The Retail Sales Data Set
Using retail-sales queries from SQL for Data Analysis to explore visualization with ggsql.
Working with DERA Notes data using Polars
Using Polars to query the SEC's DERA financial-statement notes data.
Learning Python Polars
A fast-paced introduction to Python Polars and statistics, with current setup instructions and templates.
Making plotnine more 'Pythonic'
Exploring how plotnine's grammar of graphics can fit more naturally into Python code.
Moving notes to Typst
A short account of moving the notes PDF workflow to Typst, linking the maintained collection and sources.
Solutions for 'Beaver (1968)'
Discussion and coding solutions for the Beaver (1968) chapter, using Python Polars to examine market reactions to earnings announcements.
Benchmarking local PostgreSQL-to-Parquet export paths
Comparing ways to move data from PostgreSQL into Parquet for analysis with modern dataframe tools.
Converting lazy data frames into Parquet files (Python version)
Using Python, Ibis, and db2pq to turn database queries into local Parquet data.
Converting lazy data frames into Parquet files
Moving the results of lazy R database queries into Parquet files with db2pq.
Ball and Brown (1968): Replication using Python Polars
Replicating a classic accounting study with Python Polars, and examining what its results tell us about earnings and stock returns.
Data management ideas for researchers
Practical ways to manage research data with Python, Parquet, and a local library of WRDS data.
Data management ideas for researchers (R version)
An R version of the guide to managing research data with Parquet and db2pq.
A whole research paper in Quarto
From data preparation to a PDF manuscript with a single click.
Data curation: The case of Call Reports
Building a data library from US bank Call Reports, with Polars and DuckDB.
R packages on Linux without compiling from source
A short tip for RStudio users on Linux, with current Posit Package Manager setup guidance.
Data collection (with spreadsheets)
The often overlooked work of collecting research data, and the useful role spreadsheets can play.
Working with JSON data: The case of SEC submissions data
Working with hierarchical JSON data using the SEC's filing-submissions records.
Data curation and the data science workflow
Where data curation fits into the data science workflow, illustrated with Australian stock-market data.
Some benchmarks with comp.g_secd
Comparing approaches to querying Compustat Global's daily security data, from SAS on WRDS to local Parquet files.
The best of both worlds: Using modern data frame libraries to create pandas data
Using Polars and Ibis to prepare large datasets for analysis in pandas.
Using SAS to create pandas data
Running SAS on WRDS and bringing the results into a pandas dataframe with wrds2pg.
ChatGPT as a copy editor, with Git
Using ChatGPT for feedback and Git to track changes in complex writing projects.
Shared code
What academic research might learn from open-source software about sharing and improving code.
Reproducible data collection
Making data collection reproducible, using employment data and a published research project as a case study.
2025
Responsive open-source software: Two examples from dbplyr
Two examples of how requests from users become useful features in open-source software.
Writing better SQL without writing SQL
Rebuilding an SQL query with common table expressions and dbplyr, and checking what the query actually computes.
Why aren't more data people talking about ibis?
A look at Ibis and the connection it offers between dataframe APIs and SQL databases.
But not all research is bad
Why the existence of good research does not resolve the harder problem of distinguishing it from unreliable findings.
Analysis of IPOs on the ASX
An exploration of initial public offerings on the Australian Securities Exchange.
SIRCA ASX End of Day (EOD) collection
An introduction to SIRCA's daily prices for ASX-listed securities and their use in measuring returns.
SIRCA Mergers and Acquisitions collection
Exploring SIRCA's collection of Australian mergers and acquisitions data.
farr: reusable code and teaching data
Why packaging functions and small datasets helps teaching and research, with an update to the original release advice.
Comparing skylines
Where Melbourne's skyline would fit in a published US ranking, with a reproducible calculation from historical building data.
Business valuation with ITCs
Working through Sunnova's investment tax credits, tax equity financing, and noncontrolling interests to understand value for shareholders.
Universities and willing learners
Universities, willing learners, and meal completion certificates.
Growth by industry
A rough replication of the McKinsey Valuation book's industry growth chart, with the original figure, source code, and discussion of the sample and industry mapping.
Trump's tariff formula
Unpacking the elasticity and passthrough assumptions behind the April 2025 tariff formula, combining the LinkedIn post and its mathematical explanation.
Digging into Tesla's reporting on PP&E
Reconciling Tesla's capital expenditure and PP&E disclosures, and testing explanations for an apparent discrepancy.
Accounting, finance and the bundling of education
Durable skills, job readiness, and what accounting and finance degrees offer.
Defining winter and summer in Oxford
An exploration of Oxford's seasons using weather data and Python.
Stock returns on Yahoo Finance
Examining how Yahoo Finance's closing and adjusted prices translate into stock returns.
Proposal for teaching data analytics
A proposal for learning data analytics through real data, complete code, visualisation, and core ideas from business and economics.
Retail sales
Working with Australian Bureau of Statistics retail-sales data in R.
Severance and academic incentives
Meaning, incentives, and finger traps in academic research.
2024
Getting SEC EDGAR XBRL data
Getting financial-statement and footnote data from SEC EDGAR's bulk XBRL files.
Missing Form APs?
Comparing SEC XBRL records with PCAOB data to investigate potentially missing Form AP filings.
ACNC Registry data: Arrow version
Using list columns and Parquet to represent Australian charity registry data in a single file.
Does Beardsley et al. (2021) show anything?
A simulation asks whether a reported pattern can arise without the earnings management it is claimed to reveal.
A quick look at City of Melbourne bike data
A short exploration of bicycle data from the City of Melbourne.
Working with Form AP data
Downloading and working with PCAOB Form AP data in R, illustrated by variations in engagement-partner names and audit workloads.
Defining winter and summer in Boston
Using temperature data to examine when winter and summer begin and end in Boston.
Defining winter and summer in Melbourne
Comparing calendar conventions with temperature-based definitions of Melbourne's seasons.
Defining winter and summer in Sydney
Looking at Sydney's temperature data to think about the boundaries of winter and summer.
Sunrise and sunset times
Exploring sunrise and sunset times in Melbourne and Boston.
Trading days per year (crsp.dsf)
Checking the familiar assumption of 252 trading days per year against CRSP's daily stock data.
Data visualization challenge
A plotting challenge: how should a chart represent the cumulative return on a stock?
2023
Improving performance of SQLite data
Exploring faster alternatives to SQLite for the analyses in Tidy Finance.
Calculating betas using DuckDB
Calculating stock betas with DuckDB, following the approach used in Tidy Finance.
The Gino-Colada Affair
Re-examining experimental results and the consequences of changing contested observations.
The elephant in the room: p-hacking and accounting research
P-hacking, replication, and the incentives that shape the credibility of empirical accounting research.
Adding delisting returns to monthly data
Translating SAS code into R to incorporate delisting returns in monthly CRSP data.
2022
Should Bao et al. (2020) be retracted?
Examining how corrections affect the central claims of a machine-learning study of accounting fraud.
The purpose of an accounting professor?
A response to Alex Edmans on research, institutional loyalty, collaboration, and impact in accounting academia.
2021
Do we need accounting research?
A historical argument about the costs and benefits of accounting research, its connection to teaching, and institutional demand.
Observations on accounting education in 2021
Reflections on teaching accounting, including the role of concepts, context, and judgment.
GitHub for accounting researchers
How GitHub can fit into a research workflow that preserves code and makes results easier to reproduce.
What a Journal of Finance retraction gets right
What a retraction reveals about the value of accessible data, code, and incentives to check published findings.