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

MIDAS Model

Mixed-data sampling regressions that handle indicators arriving at different frequencies inside a single forecasting equation.

概要グラフ数式参考資料

Graph workspace

MIDAS weight function with forecast decomposition

Graph workspace

The distinguishing feature of MIDAS is the parameterized lag weight function. Visualizing the weight profile w(k;θ)w(k; \theta)w(k;θ) alongside the resulting forecast decomposition - which monthly lags contribute most to the quarterly prediction - is the most direct way to understand what MIDAS does differently from bridge equations.

Chart axis metadata
xMonthly lag index kmonths back · 0-11yLag weightnormalized (sum to 1) · 0 to 0.3
MIDAS weight function w(k; theta) static chart preview showing Weight
Weighted contribution of each
high-frequency lag static chart preview showing Series 1
xMonthly lag index kmonths back · 0-11yLag weightnormalized (sum to 1) · 0 to 0.3

Current results

Peak weight lag

Explain: Peak weight lag
k∗=arg⁡max⁡kw(k;θ1,θ2)k^* = \arg\max_k w(k; \theta_1, \theta_2)k∗=argmaxk​w(k;θ1​,θ2​)

Which monthly lag receives the most weight. k∗=0k^* = 0k∗=0 means the most recent month dominates. With equal weights every lag ties and the readout reports 0.

0.0000

months back

Effective number of lags

Explain: Effective number of lags
Keff=1∑kw(k)2K_{\text{eff}} = \frac{1}{\sum_k w(k)^2}Keff​=∑k​w(k)21​

Herfindahl inverse. Equal weights: Keff=K+1K_{\text{eff}} = K+1Keff​=K+1. Concentrated weights: KeffK_{\text{eff}}Keff​ approaches 1. Measures how spread out the weight function is.

9.76

effective lags

Weighted predictor value

Explain: Weighted predictor value
zt=∑k=0Kw(k;θ)⋅xtm−kHz_t = \sum_{k=0}^{K} w(k; \boldsymbol{\theta}) \cdot x_{tm-k}^Hzt​=∑k=0K​w(k;θ)⋅xtm−kH​

The single number that summarizes the high-frequency information entering the quarterly forecast.

0.5529

same as x

Quarterly forecast

Explain: Quarterly forecast
y^tL=α+β⋅zt\hat{y}_t^L = \alpha + \beta \cdot z_ty^​tL​=α+β⋅zt​

The predicted quarterly GDP growth from the MIDAS regression.

0.63%

percent q/q

Controls

Recent months weigh more
Explain: Recent versus older months

Almon linear parameter: -0.05

Tilts the weights: negative puts more on recent months, positive puts more on older months.

Tilts the weight curve left (recent months) or right (older months).

-0.01
Explain: How the weight curve bends

Bends the weight curve: negative gives a hump if the linear tilt is positive enough, otherwise a steeper drop.

Negative with positive θ1\theta_1θ1​: weight curve peaks at an intermediate lag and declines on both sides. Positive: U-shaped, with more weight at extreme lags.

Moves the same way
Explain: Effect of the monthly indicator

MIDAS slope: 0.6

Scales how strongly the monthly indicator moves the quarterly outcome.

Larger β\betaβ: taller bars in the contribution chart. The weighted predictor has a bigger effect on the forecast.

Within one year
Explain: How many months to include

Number of HF lags: 11

How many past monthly readings to use: 2 covers one quarter, 11 covers four.

Larger K: the weight function extends over more lags. With declining weights, additional lags receive tiny weights.

0.3
Explain: Baseline forecast level

The constant term; roughly the average quarterly outcome when the weighted monthly signal is zero.

Shifts the forecast level up or down without changing the weight profile.

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