- Practice
- About 20 min
- 5 practice questions
Regression, econometrics and official statistics in India
Fit the two regression lines from means, standard deviations and r, contrast regression with correlation, and describe econometrics: its origin, definitions, objectives, methodology, error term and assumptions, and India's official statistical system.
If something looks wrong, please tell us.
By the end of this lesson you can
- Fit the regression equations of Y on X and X on Y and use them to estimate values.
- Explain what econometrics is, list its objectives and the eight steps of its methodology.
- Explain the error term U and name the bodies of India's official statistical system.
Correlation tells us how closely two variables move together. Often we want more: if X takes a given value, what is the best estimate of Y? That is regression, and when economic theory, mathematics and statistics are joined to estimate such relationships from real data, we have econometrics.
Regression
Francis Galton first used the word in 1877 while studying the heights of fathers and sons: children of unusually tall or short parents tended to 'regress', or step back, towards the average height of the population. His friend Karl Pearson confirmed the finding with more than a thousand family records. Today regression means the study of the average relationship between variables. With Y as the dependent variable and X as the independent variable, the line is the regression of Y on X, used to estimate Y for a known X.
Y on X: Y − Ȳ = r(σy ÷ σx)(X − X̄). X on Y: X − X̄ = r(σx ÷ σy)(Y − Ȳ). The two regression coefficients multiply to r².
| Correlation | Regression |
|---|---|
| Degree and direction of relationship between variables | Mathematical measure of the average relationship; 'going back' |
| Does not show cause and effect | Treats one variable as depending on the other, a functional relationship |
| Used for testing and verifying a relation | Also used to predict one value from another |
| r lies between −1 and +1 | Gives the value of the dependent variable for a known independent value |
| Can be spurious (two unrelated series moving together by chance) | The book says regression is not spurious in this way (but a regression line, like a correlation, does not by itself prove that one variable causes the other) |
| Pearson's r is confined to linear relationships | Can study linear and non-linear relationships |
Worked example 1: fitting both regression lines
Made-up data: X̄ = 20, Ȳ = 50, σx = 3, σy = 6, r = 0.6. Y on X: coefficient = r × σy ÷ σx = 0.6 × 6 ÷ 3 = 1.2. So Y − 50 = 1.2(X − 20), giving Y = 1.2X − 24 + 50, Y = 1.2X + 26. X on Y: coefficient = r × σx ÷ σy = 0.6 × 3 ÷ 6 = 0.3. So X − 20 = 0.3(Y − 50), giving X = 0.3Y − 15 + 20, X = 0.3Y + 5. Check: 1.2 × 0.3 = 0.36 = 0.6² = r², and both lines pass through the means: 1.2 × 20 + 26 = 50 and 0.3 × 50 + 5 = 20.
Worked example 2: estimating
Using Y = 1.2X + 26 from example 1, estimate Y when X = 25. Step 1: 1.2 × 25 = 30. Step 2: Y = 30 + 26 = 56. To estimate X when Y = 60, use the other line: X = 0.3 × 60 + 5 = 18 + 5 = 23. Always use Y on X to estimate Y and X on Y to estimate X.
Econometrics
According to the textbook, Irving Fisher was the first to build a mathematical equation with data, in the quantity theory of money. The Norwegian economist Ragnar Frisch named the integration of economics, mathematics and statistics 'econometrics' in 1926; he later shared the first Nobel Memorial Prize in economics, in 1969. The word comes from Greek roots meaning economy and measure, so econometrics means economic measurement: it measures economic relationships.
How econometrics is built (the book's equations)
- Economics + Mathematics = Mathematical economics
- + Statistical data and techniques = Econometrics
- Economics + Statistics + Mathematics + empirical data = Econometrics, an amalgamation of three subjects
Definitions (paraphrased): Arthur S. Goldberger: a social science applying economic theory, mathematics and statistical inference to economic phenomena. Gerhard Tintner (the book prints this name as 'Gerhard Tinbergen'): applying mathematical statistics to economic data to support mathematical-economic models with numbers. H. Theil: the empirical determination of economic laws. Ragnar Frisch: the mutual penetration of quantitative theory and statistical observation.
General objective: to give empirical content to economic theory. Specific objectives: forecasting economic phenomena; proving established relationships between variables; establishing new theories and relationships; testing hypotheses and estimating parameters.
Methodology of classical econometrics
- 1. Statement of theory or hypothesis
- 2. Mathematical model of the theory
- 3. Econometric model of the theory
- 4. Obtaining the data
- 5. Estimating the parameters
- 6. Hypothesis testing
- 7. Forecasting or prediction
- 8. Using the model for control or policy
The error term
Mathematical economics writes an exact relation, Yᵢ = β₀ + β₁Xᵢ. Econometrics adds a random (stochastic) error term: Yᵢ = β₀ + β₁Xᵢ + Uᵢ, or with several variables Y = β₀ + β₁X₁ + β₂X₂ + β₃X₃ + U. β₀ + β₁Xᵢ is the systematic (explained) part; Uᵢ is the random (unexplained) part, standing for omitted variables, so it cannot be ignored. For example, consumption depends on income, but also on habits, family size and many small influences that U captures. The linear regression model (ordinary least squares) assumes, among others, that U has zero mean, constant variance and a normal distribution; that different U values are uncorrelated; that U is independent of the explanatory variables; that explanatory variables are measured without error and are not perfectly linearly correlated; that variables are correctly aggregated and the relation correctly specified; and that the parameters are linear.
Official statistics in India
Official statistics are those published by government agencies and public bodies; they let the public judge the effect of policies. According to the textbook, the Ministry of Statistics and Programme Implementation (MOSPI) became an independent ministry in 1999, by merging the Department of Statistics and the Department of Programme Implementation, and has two wings. The Statistics wing, the National Statistical Office (NSO), comprises the Central Statistical Office (CSO) (national accounts, Annual Survey of Industries, Economic Censuses, Index of Industrial Production, consumer price indices, statistical standards), the Computer Centre and the National Sample Survey Office (NSSO), set up in 1950, the largest body conducting regular socio-economic surveys. The Programme Implementation wing covers the Twenty Point Programme, infrastructure and project monitoring, and the MP Local Area Development Scheme. There is also a National Statistical Commission and the Indian Statistical Institute, an institute of national importance.
Key takeaways: regression (Galton, 1877) estimates one variable from another: Y − Ȳ = r(σy ÷ σx)(X − X̄) and X − X̄ = r(σx ÷ σy)(Y − Ȳ). Econometrics, named by Frisch in 1926, joins economics, mathematics and statistics to measure economic relationships, following eight steps from theory to policy, and adds the error term U for omitted influences. MOSPI, with the NSO (CSO and NSSO), runs India's official statistics.
Check your understanding
-
1. The term 'regression' was first used by:
- Karl Pearson
- Isaac Newton
- Francis Galton
- Charles Spearman
Show answer
Answer: C. Galton used the term in 1877 while studying the heights of fathers and sons; Pearson later confirmed his finding.
-
2. Made-up data: X̄ = 10, Ȳ = 30, σx = 2, σy = 8, r = 0.5. The regression equation of Y on X is:
- Y = 2X + 10
- Y = 0.125X + 28.75
- Y = 2X + 30
- Y = 0.5X + 25
Show answer
Answer: A. Coefficient = r × σy ÷ σx = 0.5 × 8 ÷ 2 = 2. Y − 30 = 2(X − 10), so Y = 2X − 20 + 30 = 2X + 10. Check: at X = 10, Y = 30, the mean.
-
3. The word 'econometrics' was coined by:
- Irving Fisher
- Ragnar Frisch
- Arthur Goldberger
- Francis Galton
Show answer
Answer: B. Ragnar Frisch named the integration of economics, mathematics and statistics 'econometrics' in 1926. Fisher is credited in the book with an early data-based equation.
-
4. In Yᵢ = β₀ + β₁Xᵢ + Uᵢ, the term Uᵢ is:
- the intercept
- the slope
- the dependent variable
- the stochastic error term standing for omitted variables
Show answer
Answer: D. Uᵢ is the random or stochastic error term, the unexplained part, which represents omitted variables. β₀ is the intercept, β₁ the slope and Y the dependent variable.
-
5. Which step comes immediately after 'obtaining the data' in the methodology of econometrics?
- Estimation of the parameters of the econometric model
- Hypothesis testing
- Statement of the theory
- Forecasting
Show answer
Answer: A. The order is: theory, mathematical model, econometric model, data, estimation of parameters, hypothesis testing, forecasting, policy use. So estimation follows the data.
Sign in to keep track of the lessons you finish.