CRE: office development, through the pack contract

The same published construction schedule as the native case, declared as one cre.construction_loan contract — equity first, the facility behind it, interest on the drawn balance.

Every number below is checked against an independent reference implementation on every commit — period by period, and on each metric, inside a declared tolerance. See benchmark methodology.

The case

The same ground-up office development as the One Lincoln Street case beside it: 36 storeys in Boston's Financial District, funded quarter by quarter across the 2000–2003 build, with equity drawing first against a $110,738,000 commitment and a construction facility taking the balance once it is exhausted.

What differs is how it is said. The other case builds the funding waterfall from primitives. This one declares a single cre.construction_loan contract.

The reference

MIT OpenCourseWare 11.431J Case Assignment 3, Exhibit 7 — the same source, the same three published drivers: a sixteen-quarter draw schedule, an 8.00% rate compounded quarterly, and the net equity commitment.

Redistributable. CC BY-NC-SA 4.0. The PDF is committed once, under the native case's reference/, and both cases read the same figures from it.

What it exercises

The draw schedule stays a curve in the model rather than becoming a term. A development's funding profile is per-deal data — sixteen published quarters here, an S-curve or a contractor's schedule on the next deal — and all three are the same object. What the contract adds is the funding convention.

The curve is stated ANNUALIZED — each point is the exhibit's quarterly figure times four — and the contract divides by periods-per-year, the same convention every annual quantity in the pack follows. On this quarterly model that divides straight back to the published number. It matters because a curve is a level: a step curve returns its last point on every date, so a schedule stated as per-period totals would be correct here and would fund three times the money if the same deal were run monthly.

The result

Zero difference from the native case, in all 48 cells, and model.total, model.npv and model.wal_years agree to the last decimal. Against the exhibit itself the residuals are therefore the native case's: equity contribution and loan draw exact to the dollar in all sixteen quarters, interest within the exhibit's own rounding.

The quarter that proves the contract is 2001.4, where the commitment runs out mid-quarter and $29,430,000 of required funding splits into $10,522,000 of equity and $18,908,000 of debt. That split is stated nowhere; it falls out of the cumulative field crossing the commitment inside a period.

The delta

Nothing between the two cases. Where they ever differ, the contract is wrong: the language case is the reference, because it was validated against the exhibit first and depends on no domain vocabulary.

Against the exhibit, the deltas are inherited and unchanged — the exhibit rounds interest to whole thousands, and its stated debt-service total of 16,312,000 is the sum of those rounded quarterlies against the engine's 16,310,570 of exact ones.

Interest is paid, not capitalized, as the exhibit funds it from the equity budget as a separate line. A capitalizing facility compounds and is a different recurrence; the contract does not model it.

Run configuration

{
  "deterministic": {
    "annual_discount_rate": 0.08
  }
}

Verified results

Checked period by period: 3 series across 16 periods — 48 values in all, each within ±500.0 of the reference.

• cre.construction.equity_draw
• cre.construction.loan_draw
• cre.construction.interest

Summary metrics for the base run: