Transcription for Economics Lectures: Keeping Graphs, Models, and Numbers Straight

To transcribe economics lecture notes well, split the job by channel: let the transcript capture the professor’s reasoning — why the curve shifts, which assumption does the work, what the exam rewards — while your pen sketches every graph and your camera catches the board, then re-anchor the numbers to their labels in a same-day pass. Economics lectures sit in an awkward middle zone for speech-to-text: less notation-dense than pure math, but built on graphs the microphone can’t see and stuffed with jargon it mishears. The fix isn’t a better app. It’s knowing exactly what the transcript can and can’t carry.

Key takeaways

• Econ transcripts fail in three specific places: graph narration (“this shifts out here”) arrives as orphaned pronouns, jargon and Greek letters get homophone treatment, and numbers survive but lose their labels.
• The transcript’s real cargo is the narration — intuition, assumptions, policy stories, exam signals — which is precisely what students miss while copying diagrams.
• Sketch every graph by hand, badly and fast, with a timestamp; photograph the finished board. Ten seconds of drawing beats a paragraph of curve description.
• Do a same-day pass to reattach numbers to variables and fix the recurring jargon casualties while you still remember what was meant.
• Load model names and economist names into a custom vocabulary in week one if your tool allows it; the syllabus is the word list.

Where econ lectures break speech-to-text

An economics lecture is three interleaved streams, and transcription handles each differently.

The narration transcribes beautifully. “When the Fed raises rates, borrowing gets more expensive, investment falls, and that shifts aggregate demand” — this is ordinary English, and modern speech models (Whisper and its descendants) eat it happily. The stories, the policy examples, the intuition: near-perfect capture.

The graph narration transcribes into nonsense. Not because the words are wrong — because the meaning was never in the words. “So if this shifts out to here, the new equilibrium is down at this point” is a perfectly accurate transcript of a sentence whose entire content lived on the whiteboard. Econ is the graph discipline; a fifty-minute micro lecture can spend half its airtime pointing. This is the same silent-board problem that makes why spoken equations defeat speech-to-text required reading for STEM courses — econ just swaps the integral for a supply curve.

The jargon gets homophone treatment. Speech models are trained on everyday language, so course vocabulary drifts toward common words: elasticity terms scramble, *ceteris paribus* arrives phonetically creative, Greek letters become names, “IS-LM” becomes “is LM,” and Cobb-Douglas gets spellings no economist would recognize. The errors are annoying but consistent — the same term breaks the same way all semester, which matters for the fix below.

And the numbers half-survive. Digits transcribe well: “point seven” reliably becomes 0.7. What evaporates is the attachment — *which* 0.7? The marginal propensity to consume, an elasticity, a made-up example value? The transcript keeps the number and drops the label, which is worse than losing both, because an unattached number looks trustworthy.

The capture split: what goes where

So run three channels, each doing what it’s good at:

• Mic → reasoning. Record the lecture; let the transcript own the explanations, assumptions, policy stories, and the “this is the kind of thing I’d ask on the midterm” asides.
• Pen → graphs. Every time a diagram goes up, sketch it — ten seconds, ugly, two crossing lines and an arrow for the shift. Hand-drawing the shift is not overhead; drawing the mechanism is how the mechanism sticks. Write the time (or slide number) next to each sketch.
• Camera → the finished board. Photograph at natural pauses, not mid-derivation. The photo carries the clean version; your sketch carries the motion.

The timestamps are the weld. A transcript with timestamps, sketches with times in the margin, and photos with capture times zip back together in minutes. Without them you’re matching “this curve here” against six photos of curves.

This is the identical division of labor that works one building over — the engineering-lecture version of this problem swaps demand curves for free-body diagrams and changes nothing else.

The same-day pass: how to transcribe economics lecture notes into something a problem set can use

The raw transcript plus sketches plus photos becomes actual notes in one twenty-minute pass, done the same day while context is fresh. Cornell’s note-taking system has insisted on the review-soon step for decades; for econ transcripts it has three specific jobs:

1. Reattach the numbers. Everywhere a bare value sits in the transcript, label it: “0.7” becomes “MPC = 0.7 (assumed, for the multiplier example).” Today this takes seconds; at problem-set time it’s guesswork, and guessed labels produce confidently wrong answers.
2. Fix the jargon casualties. The mistranscriptions are consistent, so they’re bulk-fixable — search for the usual suspects and correct them. Better: prevent them. If your tool supports it, load the course’s model names, economist names, and abbreviations straight from the syllabus — teaching your transcriber the course jargon takes ten minutes in week one and pays all semester.
3. Marry sketches to text. Drop each sketch (or board photo) into the notes at its timestamp, so the professor’s explanation sits next to the diagram it was explaining. This is the moment the orphaned pronouns get their referents back — “this shifts out” finally means something because the arrow is right there.

While you’re in the file, star the exam signals. Econ professors telegraph constantly — “people always shift the wrong curve here,” “I don’t care about the derivation, I care whether you know when it applies.” That narration is the highest-value content in the room, and it’s exactly what students lose while racing to copy a diagram they could have photographed.

One boundary before the essays

Econ courses also assign writing — policy memos, short essays, discussion posts — and a semester of clean transcripts makes a tempting quarry. Quote your professor’s framing sparingly and cite it; and don’t paste transcript passages through an AI rewrite into submitted work. The result isn’t your analysis, and AI-processed prose carries a statistical fingerprint that detection tools are built to catch. Draft from your own understanding; if AI helped along the way, run it through a detector before you submit (compare plans if it becomes a habit). More workflows live in more of our transcription guides.

Frequently asked questions

What does transcription get wrong in economics lectures specifically? Three things. Graph narration arrives as orphaned pronouns — “this curve shifts out here” is meaningless without the board. Homophones and jargon get mangled: elasticity terms, Greek letters, and phrases like ceteris paribus come out as ordinary-word soup. And numbers, which transcribe accurately as digits, lose their labels — you get “0.7” faithfully but not whether it was the marginal propensity to consume or an elasticity. The transcript is excellent on reasoning and stories, weak on anything that lived on an axis.

Are spoken numbers reliable in a transcript? The digits themselves usually survive — modern speech models handle “point seven” and “three percent” well. What gets lost is the attachment: which variable the number belongs to, whether it was an assumption or a result, and whether the professor said it was made up for the example. In a same-day pass, re-anchor every number to its label while you remember the context. A transcript full of accurate but unattached numbers is a trap during problem sets.

How do I capture the graphs if the transcript can’t? Sketch them live — badly is fine — and photograph the board at natural pauses. A supply-and-demand diagram takes ten seconds to draw, and hand-drawing the shift is itself how the mechanism sticks. Then note the time or slide number next to your sketch so it reconnects to the transcript afterward. The transcript carries the professor’s explanation of why the curve shifts; your sketch carries which curve and which direction. Neither channel works alone.

Do model names like IS-LM and Solow transcribe correctly? Inconsistently. Named models arrive as their phonetic neighbors — IS-LM becomes “is LM” or worse, Cobb-Douglas gets creative spellings, and abbreviations spoken as letters scatter into words. If your tool supports a custom vocabulary, load the course’s model names, economist names, and abbreviations from the syllabus in week one. If not, learn the recurring mistranscriptions and fix them in your cleanup pass — the errors are at least consistent, which makes them searchable and correctable in bulk.

Is recording worth it for a problem-set-based econ course? Yes, but for the setup, not the algebra. Lectures in problem-heavy courses contain the framing that problem sets silently assume: what the model is for, which assumptions do the work, what breaks when they fail, and the professor’s asides about what the exam actually rewards. That narration rarely makes it into anyone’s notes because students are busy copying derivations. Let the transcript hold the reasoning and the warnings; work the algebra from the textbook and your board photos.

The bottom line

Economics lectures split their meaning across the professor’s mouth and the professor’s board, and the microphone only reaches one of them. Stop asking the transcript to describe curves — it can’t, and trying produces pronoun soup. Point it at the narration: the assumptions, the intuition, the policy stories, the exam telegraphs. Sketch the graphs with your own hand, photograph the clean board, and spend twenty same-day minutes welding the channels together and reattaching numbers to their labels. The result is the set of notes econ students rarely have: the diagrams *and* the reasoning that makes them move.