How to Know When to Build (and When to Wait): Predicting Real Estate Cycles

The worst time to start building apartments is when everyone else is building apartments.

That sounds obvious, but real estate development operates on 18-24 month timelines. By the time you see cranes in the air across town, your project is already approved, financed, and under construction. You can’t easily back out.

The key is predicting the cycle before it’s visible.

The Problem: Supply Lags Demand (Until It Doesn’t)

Real estate markets oscillate between undersupply and oversupply:

Phase 1: Undersupply

Rents rise as demand outpaces supply
Occupancy stays high (95%+)
Developers see the opportunity and start planning projects

Phase 2: Construction boom

Multiple projects break ground simultaneously (each started when market looked hot)
Supply pipeline swells but takes 18-24 months to deliver
Rents continue rising because new units haven’t hit the market yet

Phase 3: Oversupply

All those projects deliver at once
Vacancy spikes, rents stagnate or fall
Developers who started projects 2 years ago are now stuck in a soft market

Phase 4: Correction

New construction stops (nobody wants to start projects in a soft market)
Demand slowly absorbs excess supply
3-5 years later, the market is tight again and the cycle repeats

The question: How do you know when you’re in Phase 1 (go build) vs. Phase 2 (market looks hot but you’re late)?

What We Built: A Forecasting Model for Supply-Demand Dynamics

We built a model that ingests 30+ years of historical data to forecast where the market is in the cycle—and where it’s headed.

Inputs:

Historical deliveries: New units added to the market each quarter (from CoStar, Reis, or local permitting data)
Absorption rates: Net units absorbed by renters each quarter
Rent growth: Year-over-year change in effective rents
Pipeline data: Projects under construction + projects planned (not yet started)

The model does two things:

1. Estimates current market state (even when data is incomplete) Using a Kalman filter (a state-space estimation technique), the model infers:

True underlying demand (smoothing out noise in quarterly absorption data)
Supply-demand balance (are we currently tight or loose?)
Rent growth momentum (is growth accelerating or decelerating?)

This matters because raw data is noisy. One quarter’s absorption spike might be statistical noise, not a real trend. The model filters out noise to reveal the underlying pattern.

2. Forecasts the next 8 quarters Based on known pipeline (units under construction) and historical demand patterns, the model projects:

When will oversupply hit?
How severe will it be?
How long until the market rebalances?

Output: A clear signal: “Market is entering oversupply in Q3 2026. Delay new starts until Q2 2027.”

Or: “Market remains tight through 2026. Green light to break ground.”

What This Means for Developers

Better timing: Start projects when the market is tight and deliver into continued demand—not into a glut.

Avoided losses: One bad development cycle can wipe out years of profits. Avoiding one mistimed project pays for decades of forecasting.

Competitive advantage: While competitors build when the market looks hot (lagging indicator), you build when the model says it’s still early (leading indicator).

Confidence in go/no-go decisions: Instead of gut-feel or “the market feels strong,” you have data-backed conviction.

The Science Behind It

This isn’t a black-box AI predicting rents. It’s a state-space model that treats supply-demand dynamics as a system with inertia:

Demand evolves based on job growth, household formation, and migration (slow-moving, predictable)
Supply responds to demand with an 18-24 month lag (pipeline data tells us what’s coming)
Rent growth is the observable output that reflects the underlying supply-demand balance

The Kalman filter uses observed rent growth and absorption data to infer the unobservable true demand state—then projects forward using known pipeline.

Why this works: Supply is predictable (we know what’s under construction). Demand is relatively stable (metro-level job growth doesn’t swing wildly quarter-to-quarter). The uncertainty is in timing the transition from tight to loose market—and that’s exactly what the model is designed to estimate.

Who This Works For

This approach makes sense if you’re:

Developing multifamily or commercial properties and timing is everything
Acquiring land for future development and need to know when to pull the trigger
Raising capital for new projects and need data to justify your market timing thesis
Operating in volatile markets (fast-growing Sunbelt cities) where cycles are sharp and painful

The ROI Calculation

Scenario: You’re deciding whether to start a 200-unit project. Construction cost = $40M, expected stabilized NOI = $3M/year (7.5% yield).

Mistimed project (deliver into oversupply):

Lease-up takes 18 months instead of 9 months
Stabilized occupancy = 88% instead of 95%
Year 1 NOI = $2.3M instead of $3M
Lost value: $10M+ in lower NOI and longer time to stabilization

Well-timed project:

Deliver into tight market, lease-up in 6 months, stabilize at 96% occupancy
Full projected returns achieved

Avoiding one mistimed project = $10M+ in preserved value.

The forecasting model costs a fraction of that.

What We Learned Building This

1. Data quality beats model complexity Early versions used fancy machine learning. Didn’t work. The breakthrough came from using a simpler model (Kalman filter) with clean, long-term historical data (30 years of deliveries and absorption).

2. Pipeline data is the most predictive variable Rent growth today doesn’t tell you much about rent growth 2 years from now. But units under construction today? That’s exactly what will hit the market in 18-24 months. Pipeline is destiny.

3. Developers trust models that match their intuition (most of the time) When the model says “market is tight, build now” and developers agree, it builds trust. When the model says “market looks hot but you’re late,” that’s when it provides real value—catching what intuition misses.

How It Works in Practice

Quarterly workflow:

Update data (15 min): Import latest CoStar/Reis deliveries, absorption, rent data
Run model (instant): Kalman filter estimates current state, forecasts 8 quarters ahead
Review forecast (10 min): Check model output against market intelligence (what are competitors saying? what’s local media reporting?)
Make decision: Green light new projects, pause pipeline, or adjust underwriting assumptions

Total time: 30 minutes per quarter to stay ahead of the cycle.

Beyond Timing: Strategic Use Cases

Once you have a calibrated supply-demand model, you can answer bigger questions:

Land acquisition: Is this site better suited for immediate development or land-banking for 2-3 years?

Capital allocation: Which markets are early-cycle (invest now) vs. late-cycle (harvest and redeploy elsewhere)?

Underwriting assumptions: Should we assume 5% rent growth or 2% in our pro forma? The model provides a data-backed answer.

Investor communication: Explain to LPs why you’re pausing development in a “hot” market (because the model shows oversupply coming) with data, not gut-feel.

Next Steps

If you’re making multimillion-dollar development timing decisions based on intuition or anecdotal market intel, here’s where to start:

Gather historical data: 10+ years of deliveries, absorption, rent growth for your market
Map the current pipeline: What’s under construction? What’s planned but not yet started?
Identify your biggest unknown: Is it demand trends? Supply timing? Rent growth trajectory?

Then ask: Could a forecasting model reduce timing risk and improve returns?

That’s the question we answered for this developer. The result: confidence in go/no-go decisions, backed by 30 years of data, not 30 days of gut-feel.

Developing in a cyclical market and want to see how this could work for your pipeline? Let’s talk about building a custom supply-demand forecasting model for your target markets.