メインコンテンツへ移動

Macro by Mark の Cookie.ログイン、言語、安全性のために必須 Cookie を使用します。許可いただいた場合は、設定の保存、分析、Meta(Facebook)ピクセルによる広告計測のために任意の Cookie も使用します。詳しくはCookie ポリシーをご覧ください。

Macro by MarkMacro by Mark
  • ホーム
読み込み中…
読み込み中…
← モデル
Macro概念政策モデル

Bridge equations Model

Simple regressions that bridge from monthly indicators to a quarterly target like GDP, refreshed as each indicator lands.

概要グラフ数式参考資料

Understand the mathematics

Follow the same equations and assumptions one section at a time or in a continuous reading.

Hover or focus a symbol defined in this model to find its exact matches. Select it to keep the highlight; Escape clears the selection.

Temporal Aggregation and the Bridge Regression

GDP is observed quarterly. Monthly indicators are observed monthly. The bridge equation operates on the quarterly frequency: it regresses quarterly GDP on quarterly aggregates of monthly indicators. The aggregation step - converting monthly to quarterly - is not an afterthought; it defines the information content available to the bridge.

For flow variables (production, sales, spending), the quarterly value equals the sum (or equivalently the mean times 3) of the three monthly values within the quarter. For stock variables (employment, inventories), the quarterly value equals the end-of-quarter (month 3) observation. Mixing these conventions - using the average of a stock variable, or the end-of-quarter value of a flow - introduces a systematic bias in the bridge coefficient.

xj,Qflow=13(xj,m1+xj,m2+xj,m3)x_{j,Q}^{\text{flow}} = \frac{1}{3}(x_{j,m_1} + x_{j,m_2} + x_{j,m_3})xj,Qflow​=31​(xj,m1​​+xj,m2​​+xj,m3​​)

Temporal aggregation for flow variables. The quarterly average equals the arithmetic mean of the three monthly values.

xj,Qstock=xj,m3x_{j,Q}^{\text{stock}} = x_{j,m_3}xj,Qstock​=xj,m3​​

Temporal aggregation for stock variables. Only the end-of-quarter value matters.

yQ=c+∑j=1kβjxj,Q+eQ,eQ∼(0,σe2)y_Q = c + \sum_{j=1}^{k} \beta_j x_{j,Q} + e_Q, \quad e_Q \sim (0, \sigma_e^2)yQ​=c+j=1∑k​βj​xj,Q​+eQ​,eQ​∼(0,σe2​)
yQy_QyQ​cccxj,Qx_{j,Q}xj,Q​

The core bridge regression. Each coefficient measures the marginal predictive power of indicator j for GDP.

y^Q(v)=c^+∑j=1kβ^jx^j,Q(v)\hat{y}_Q(v) = \hat{c} + \sum_{j=1}^{k} \hat{\beta}_j \hat{x}_{j,Q}(v)y^​Q​(v)=c^+j=1∑k​β^​j​x^j,Q​(v)
vvv

The nowcast at vintage vvv. As more monthly data arrives, x^\hat{x}x^ approaches xactualx_{\text{actual}}xactual​ and the nowcast converges to the fitted value.

Classify each indicator as flow or stock

The aggregation method must match the national accounts convention for the quarterly target
Flow: IP, retail sales, exports. Stock: employment, inventories. Survey: PMI (use average).\text{Flow: IP, retail sales, exports. Stock: employment, inventories. Survey: PMI (use average).}Flow: IP, retail sales, exports. Stock: employment, inventories. Survey: PMI (use average).

Aggregate observed monthly values to the quarterly frequency

The bridge regression operates on quarterly data
xj,Q=13∑m∈Qxj,m(for flows)x_{j,Q} = \frac{1}{3}\sum_{m \in Q} x_{j,m} \quad \text{(for flows)}xj,Q​=31​m∈Q∑​xj,m​(for flows)
xj,Qx_{j,Q}xj,Q​xj,mx_{j,m}xj,m​

Estimate the bridge regression by OLS on the historical quarterly sample

OLS gives the best linear unbiased predictor under standard assumptions
β^=(XQ′XQ)−1XQ′yQ\hat{\boldsymbol{\beta}} = (\mathbf{X}_Q'\mathbf{X}_Q)^{-1}\mathbf{X}_Q'\mathbf{y}_Qβ^​=(XQ′​XQ​)−1XQ′​yQ​

For the current quarter, fill missing monthly values with AR forecasts

The ragged edge means not all months are observed
x^j,m=ϕ^jxj,m−1+c^jfor each missing month m\hat{x}_{j,m} = \hat{\phi}_j x_{j,m-1} + \hat{c}_j \quad \text{for each missing month } mx^j,m​=ϕ^​j​xj,m−1​+c^j​for each missing month m
x^j,m\hat{x}_{j,m}x^j,m​

Compute the quarterly aggregate using observed + filled values

The bridge regression requires a complete quarterly observation
x^j,Q(v)=13(xj,m1obs/fill+xj,m2obs/fill+xj,m3obs/fill)\hat{x}_{j,Q}(v) = \frac{1}{3}(x_{j,m_1}^{\text{obs/fill}} + x_{j,m_2}^{\text{obs/fill}} + x_{j,m_3}^{\text{obs/fill}})x^j,Q​(v)=31​(xj,m1​obs/fill​+xj,m2​obs/fill​+xj,m3​obs/fill​)
vvv

Plug into the bridge equation to produce the nowcast

The fitted value from the estimated bridge equation is the GDP nowcast
y^Q(v)=c^+β^1x^1,Q(v)+⋯+β^kx^k,Q(v)\hat{y}_Q(v) = \hat{c} + \hat{\beta}_1 \hat{x}_{1,Q}(v) + \cdots + \hat{\beta}_k \hat{x}_{k,Q}(v)y^​Q​(v)=c^+β^​1​x^1,Q​(v)+⋯+β^​k​x^k,Q​(v)
vvv

Why not use monthly data directly?

MIDAS regression (Ghysels, Santa-Clara, Valkanov 2004) does exactly this: it regresses the quarterly target on monthly lags without pre-aggregation. Bridge equations lose information by averaging within the quarter. MIDAS preserves the within-quarter timing. The trade-off: MIDAS requires more parameters (lag polynomial specification) and is harder to implement in real-time with a ragged edge.

Notations

vvv

Information vintage

QQQ

Quarter index

m1,m2,m3m_1, m_2, m_3m1​,m2​,m3​

Months of the quarter

About these symbols

Information vintage: Information vintage: the point in time (within the quarter) at which the nowcast is computed

Quarter index: Quarter index

Months of the quarter: The three months within quarter Q

    Transition checks

    Forecast origin
    y^T+h∣T=ET[yT+h],h=1,2,…\widehat{y}_{T+h|T}=\mathbb{E}_T[y_{T+h}],\qquad h=1,2,\ldotsy​T+h∣T​=ET​[yT+h​],h=1,2,…

    Conditioning on information at T separates the sample from the projection. The forecast can be smooth even when later realized data will not be.

    The bridge point is shared so the eye can see the handoff. After that point, the red line is a model forecast, not observed history.

    Do not score the forecast segment as realized data or read it as a policy promise.

    Uncertainty propagation
    Var⁡T(yT+h)=σ2∑j=0h−1ψj2\operatorname{Var}_T(y_{T+h})=\sigma^2\sum_{j=0}^{h-1}\psi_j^2VarT​(yT+h​)=σ2j=0∑h−1​ψj2​

    Each future innovation contributes through its impulse weight ψj\psi_jψj​. Unless the future is degenerate, longer horizons carry more conditional variance.

    The band widens when shocks accumulate through the model recursion.

    Do not treat the band edge as an alternative deterministic scenario.

    Variable timing

    Observed history

    yt∈ITy_t\in\mathcal{I}_Tyt​∈IT​

    The history segment is realized data available at the forecast origin.

    Conditional mean

    y^T+h∣T=E[yT+h∣IT]\widehat{y}_{T+h|T}=\mathbb{E}[y_{T+h}\mid\mathcal{I}_T]y​T+h∣T​=E[yT+h​∣IT​]

    The forecast path is a conditional expectation, so it can be smooth while future realizations remain volatile.

    Prediction band

    σh2=Var⁡(yT+h∣IT)\sigma_h^2=\operatorname{Var}(y_{T+h}\mid\mathcal{I}_T)σh2​=Var(yT+h​∣IT​)

    Band width is horizon-specific uncertainty, not a second deterministic transition.

    Keep going

    Read next

    Continue with a related concept, model, or example.

    Theory model

    IS-LM Model

    a well-connected theoretical model two steps from here

    New angle: theoretical model →
    Macro by Mark
    ヘルプ
    プライバシーポリシー利用規約倫理とコンプライアンス

    © 2026 Mark Jayson Nation LLC