supply_demand_var#

  • Series: lecture-python-advanced.myst

  • File: lectures/supply_demand_var.md

  • Audit date: 2026-08-26

  • Corpus snapshot: b83d6da399

  • Categories audited: writing, math, code, figures, references, links, admonitions (JAX out of scope)

  • Overall score: 8.6 / 10

  • Priority: NONE

Score breakdown#

Category

Score

One-line note

Writing

8/10

qe-writing-002 ×3; qe-writing-007 ×2.

Math

8.5/10

qe-math-002 ×1.

Code

7.5/10

qe-code-001 ×9.

JAX

out of scope

JAX rules target lecture-jax.

Figures

6/10

qe-fig-003 ×4; qe-fig-005 ×3; qe-fig-008 ×8, +1 more.

References

10/10

no mechanical violations detected.

Links

10/10

no mechanical violations detected.

Admonitions

10/10

no mechanical violations detected.

Issues#

Critical#

None found.

High severity#

  • [qe-code-001] (reviewer) — Follow PEP8 unless closer to mathematical notation. Count: 9. Lines: 614, 618, 765, 803, 844, 480, 495, 1158, 1214. Example: compound statements are used systematically to build the state-space blocks - B = np.zeros((n, 1)); B[0, 0] = 1.0 at 614 and the same pattern at 615, 620, 621, 765-770, 797, 799, 800, 805, 806 and 844, seventeen lines in all - and 618 and 803 pack three statements including a return onto one line (e = np.zeros(n); e[k - 1] = 1.0; return e), which PEP8 asks to be split. Top-level defs are separated by a single blank line rather than two throughout (480, 488, 495, 505, 592, and the same at 439, 669, 761, 792, 1158), while the one-line docstrings at 481, 489, 496, 506, 587, 670, 762, 793 and 1159 use single double-quotes against the triple quotes of the multi-line docstring at 593-601. make_ar_price at 1158 declares a parameter nX_extra=0 that its body never uses. And 1214 rebinds the module-level scale set at 514, inside an exercise solution, so a reader who runs the exercises and then re-runs an earlier cell gets different numbers.

  • [qe-fig-008] — Use lw=2 for line charts. Count: 8. Lines: 529, 530, 697, 699, 1039, 1040, 1046, 1047. Example: plot() without lw=.

  • [qe-math-002] — Use \top for transpose notation. Count: 1. Lines: 975. Example: ^T transpose in )^T.

Medium severity#

  • [qe-fig-001] — Do not set figure size unless necessary. Count: 3. Lines: 527, 695, 1038. Example: figsize=.

  • [qe-fig-003] — No matplotlib embedded titles. Count: 4. Lines: 532, 706, 1041, 1048. Example: .set_title.

  • [qe-fig-005] — Descriptive figure names for cross-referencing. Count: 3. Lines: 526, 694, 1026. Example: code-cell figure without mystnb figure metadata.

  • [qe-writing-002] (reviewer) — Keep writing clear, concise, and valuable. Count: 3. Lines: 890, 1104, 1110. Example: the lecture’s two conclusions are each stated three or four times, twice in near-identical words. “Today’s suppliers and today’s demanders both react to supply and demand shocks, through their common effect on the prospective path of prices” (891-893) reappears 210 lines later as “Today’s suppliers and today’s demanders each react to supply and demand shocks, through their common effect on the expected future path of prices” (1104-1105), and the same point was already made in full in its own subsection at 321-334. The shock-invertibility conclusion likewise appears at 948-952, at 1064-1065, again in the {admonition} at 1067-1078, and a fourth time at 1112-1118 - where 1110 introduces it as “Finally, an aside” although it is one of the four things the Overview promises at 40-43. A Summary that restates is legitimate; three of these four are not in the Summary.

  • [qe-writing-007] (reviewer) — Use visual elements to enhance understanding. Count: 2. Lines: 447, 966. Example: the lecture computes roots four times and never draws them. The supplier’s two roots (203-204), the demander’s eight (278-281) and the market operator’s eight (447-448) are the geometric heart of the whole construction - four reciprocal pairs \((\lambda_i, \beta/\lambda_i)\) split by the circle of radius \(\sqrt\beta\) - and every one of them reaches the reader only as a printed list of moduli (206-208, 283-284, 457). One complex-plane scatter with the unit circle and the \(\sqrt\beta\) circle drawn would make “collect the four stable roots” (253, 387) visible instead of arithmetic. Second, and more consequential: 963-966 asserts that “some of the zeros do lie inside” the unit circle, which is the single fact the entire second half of the lecture rests on, and the zeros of \(G_z(I - A_F L)^{-1}C\) are never computed, printed or plotted - the reader is asked to take the invertibility failure on trust and then shown its consequences at 988-1053.

Low severity#

None found.

Strengths#

  • The Overview commits to four deliverables as a bulleted list (31-43) and the lecture delivers them in that order and says so as it goes - the two dynamic curves at 167-334, the Big-\(X\)-little-\(x\) recasting at 719-756, and the vector-autoregression difficulty at 897-1078 - with 45-46 naming the organising distinction (representations outside and inside a rational expectations equilibrium) that the Summary at 1085-1095 then closes on.

  • The solution method is stated once, as an indented block quote, before it is used three times: 175-177 says factor the characteristic operator, solve the stable root backwards into a feedback on lags and the unstable root forwards into a geometric sum - and then that recipe is visibly applied to \(\phi_s\) (183-238), to \(\phi_d\) (242-319) and to \(\psi\) (381-428) with the same vocabulary each time.

  • The market Euler equation is derived in four explicitly numbered steps (348, 367, 385, 402), and the one step that could look like a trick is flagged as temporary and then undone: Step 1 says “temporarily erase the operator \(E_t\)” and Step 4 is titled “reinstate \(E_t\) in front of the continuation shocks”, so the reader is never left wondering where the expectation went.

  • The same equilibrium is computed twice by genuinely different routes and the agreement is checked numerically rather than asserted: 683-686 prints the maximum discrepancy in both the feedback coefficients and the impulse responses, 689-690 explains the residual (the ridge at 628), and 694-711 overlays the two sets of responses as lines and markers so the reader can see them coincide.

  • The Big-\(X\)-little-\(x\) fixed point is verified rather than described: 841-849 evaluates the supplier’s price-taking best response on the equilibrium state and prints its distance from the equilibrium rule, and exercise 3 (1239-1268) asks the reader to repeat the check for the demander with its four own lags.

  • The exercises test the two claims a sceptical reader would want tested. Exercise 1 (1122-1174) checks that the own-lag feedback \(\delta_s/\beta\) is invariant to the price process by feeding the regulator three AR(1) prices, and explains the invariance from certainty equivalence before the code runs. Exercise 2 (1183-1225) sweeps \(g_s \in \{0.1, 1, 10, 50, 100\}\) and measures the fraction of the supply-shock response arriving at lag zero, with the direction predicted at 1203-1211 and the mechanism confirmed at 1227-1234.

  • Two paragraphs at 149-152 name the one parameter choice that drives the entire second half - \(g_s = 10\) is a hundred times \(g_d = 0.1\), so quantity adjusts sluggishly - and 1058-1062 then reads that same asymmetry off the whitener’s impulse responses, so the numerical result at 1046-1051 is traced back to a parameter the reader was warned about 900 lines earlier.

  • The {admonition} at 1067-1078 states the moral in a form that generalises past this model: a vector autoregression always recovers some fundamental noise for the data, and Sims’s accounting always produces a tidy decomposition, but the data’s fundamental noise need not be the agents’ - which is the reusable lesson rather than a summary of the example.