Difference Between Semi-Log Model and Log-Linear Model (With Examples)

Semi-log Model (also called semi-logarithmic model) involves taking the natural logarithm of only one of the variables (either the dependent or the independent), while the other remains in its original (level) form. This produces a model that is still linear in parameters and can be estimated by ordinary least squares (OLS), but the economic interpretation mixes percentage and absolute changes.

There are two common subtypes:

  1. Log-lin model (log of dependent variable, linear independent variable):
    ln(Y) = β₀ + β₁X + ε
    • Interpretation: A one-unit increase in X is associated with an approximate 100 × β₁ % change in Y (holding other factors constant).
    • Useful when Y grows (or declines) at a constant percentage rate with respect to X (e.g., growth models, wages, prices over time).
    • Example: Modeling wages as a function of experience
      ln(Wage) = 2.5 + 0.08 × Experience
      Each additional year of experience is associated with an ≈ 8% increase in wage.
  2. Lin-log model (linear dependent variable, log of independent variable):
    Y = β₀ + β₁ ln(X) + ε
    • Interpretation: A 1% increase in X is associated with a β₁ / 100 unit change in Y.
    • Useful when the effect of X on Y diminishes as X gets larger (diminishing returns).
    • Example: Modeling household food expenditure as a function of income
      FoodExp = 500 + 120 × ln(Income)
      A 1% increase in income is associated with a 1.20 unit increase in food expenditure.

Log-linear Model (also called double-log, log-log, or constant-elasticity model) takes the natural logarithm of both the dependent and independent variables:
ln(Y) = β₀ + β₁ ln(X) + ε
(or equivalently, Y = e^{β₀} × X^{β₁})

  • Interpretation: β₁ is the elasticity — a 1% increase in X is associated with a β₁ % change in Y (constant elasticity).
  • Useful when you want to estimate constant proportional responses (e.g., demand functions, production functions).
  • Example: Modeling quantity demanded as a function of price
    ln(Quantity) = 4.2 – 1.5 × ln(Price)
    The price elasticity of demand is –1.5 — a 1% increase in price is associated with a 1.5% decrease in quantity demanded.

Key Distinctions

AspectSemi-log ModelLog-linear Model
Logs usedOnly one variable (Y or X)Both Y and X
Interpretation of β% change in one variable vs. absolute change in the other% change in Y vs. % change in X (elasticity)
Typical useGrowth rates, diminishing returnsConstant elasticity relationships
Plot appearanceStraight line on semi-log paper (one axis logged)Straight line on log-log paper (both axes logged)
Cannot use withNon-positive values in the logged variableNon-positive values in either variable

Both models are linear in parameters, so OLS works directly, but you must be careful when comparing R² or making predictions (e.g., for log-lin or log-linear, you often need to adjust for the retransformation bias when predicting the level of Y).

These functional forms let you capture non-linear relationships in the original variables while keeping the estimation simple and the coefficients economically meaningful.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *