• 中文version喺英文之後

‼️ The Big Picture Framework

1. Formalisation = Frame the question
2. Estimation = Use data to produce evidence
3. Inference = Decide what the evidence actually tells us

Econometrics always circles back to the same three fundamental building blocks

• No matter how complex the method, it's always these three steps

⭐️ Formalisation

• What it is: Turning a broad question into something precise and testable
• The Process:

  • Start with a real-world question: "Does education matter for wages?"

  • Define your variables:

    • What we want to explain (outcome variable)
    • What might explain it (explanatory factors)

  • Choose your modelling approach:

    • Many possible ways to describe the relation
    • Economists happen to love linear models (e.g., $y_i = \beta_0 + \beta_1 x_i + u_i$)

  • Check whether we can Identify the parameter

    👈 might be difficult to understand, but we will further elaborate

    • ✅ REMEMBER THE IDEA FOR NOW: "Given infinite data from the population, can we uniquely determine the parameter values?"

  • State the required assumptions:

    • What conditions must hold for our method to give reliable results?
    • Each technique has its own set of assumptions we must be willing to accept
    • Think about whether those assumptions match up with reality or whether we can justify them

⭐️ Estimation

• What it is: find “a-number” closest to the true parameter value
• Properties of a Good Estimator:
  • Unbiased: Not systematically too high or too low
  • Efficient: Makes best use of available information
  • Robust: Reliable even if world isn't exactly as assumed 👉 This is where the question moves from "story" to evidence

⭐️ Inference

• What it is: Having “a-number” isn't enough — we need to interpret and evaluate them
• Key Questions:
  • "Is this effect real, or just noise?"
  • Tests: Are relationships strong enough to matter?
  • Precision: Report confidence ranges, not just point estimates
  • Robustness: Do conclusions hold with different data slices or model tweaks?
  • Most importantly: Distinguish correlation from causation
    • Modern econometrics cares deeply about cause-and-effect  ⚠️ 2nd Year econometrics will not teach this. You will only learn correlation NOT causation