TOPIC1A: INTRODUCTION

1.1 Review of Econometrics Concepts

1.1.1 What is econometrics?
• Various authors:

(a) Gujarati/Maddala: Econo/ metrics means “economic measurement” [Gujarati]

o Involves the application of statistical and mathematical methods to analyze economic data with the aim
of empirically testing economic theories
o Combines economic theory (micro & macro econ), mathematical economics, economic statistics, and
mathematical statistics

(b) Greene: Econometrics is the field of economics that concerns itself with the application of mathematical
statistics and the tools of statistical inference to the empirical measurement of relationships postulated by
economic theory

(c) Cowles Commission (1940?)

• Econometrics is the statistical study of the interaction of rational decision makers
• Econometrics is empirical measurement of economic phenomena & the application of statistical inference
tools

(d) Hendry (1980):

• Econometrics is an analysis of the relationships between economic variables (such as quantities and prices,
incomes and expenditures, etc.) by abstracting the main phenomena of interest and stating theories thereof in
mathematical form

SUMMARY:
• Measurement of economic phenomena
• Application of mathematics and statistics
• Based on economic theory
• “Epistemology” → the theory of knowledge in terms of methods, validity, and scope, and the distinction
between justified belief and opinion

1
1.1.2 Model
[Link] What is a “model”?
• A simplification of real-world phenomena but not the real thing itself
o Real-world phenomena?
▪ A system or process
• System may be
o Physical – physics & chemistry
o Biological
o Social
• It is an abstraction of reality
o it takes out the complexity of the real world
o isolates important phenomena from unimportant ones for study

• We use models as tools for

o understanding the real/natural world – by simplifying that real/natural world
o making predictions about the phenomenon at hand/of interest → e.g., mathematical &
statistical models

• A model is a theory spelled out in fully testable terms

• It includes a system of “hypotheses”

What is a “hypothesis”?
• It is a statement of an expected (a hunch or believe) and testable relation among variables (See later)

So,
• Models intermediate between theory and data
• Data are used to confirm or falsify theories and models (See later)

• A model can be a representation of objects, systems or events, processes, etc

o In social sciences ➔ modeling/simulating/mimicking the behavior of human agents – people,
societies, social systems, etc [individuals, farms, firms, economies, govt]
o In physical sciences ➔ modeling/simulating/mimicking the behavior of physical
agents/elements – matter [physics], chemical [chemistry], bio [biology] & their systems, etc

• In this course
o Human behavior
▪ Difficult to predict
E.g., a “random walk”

2
 ▪ Occurs within contexts

• So, what do you want to be when you grow up?

3
[Link] Issues in modeling
[Link].1 To be useful, a model MUST strike a balance between realism and manageability/tractability
▪ It should be simple and easy to understand
▪ However, it should not be oversimplified or driven by unrealistic assumptions
• Should yield sufficiently accurate predictions based on economic theory
• I.e., it should be parsimonious (not too small but also not too big/large)
o Parsimony = universal scientific principle that no extraneous causes or
forces should be invoked in explanation of a phenomenon than are
minimally required to account for the data

Occam's Razor – is the problem-solving principle that recommends searching for explanations constructed with the
smallest possible set of elements. It is also known as the principle of parsimony or the law of parsimony

[Due to William of Ockham, a 14th-century English philosopher and theologian]

[Link].2 Modeling is both an art as well as a science

o Art: Need skills to be able to balance between realism and manageability/tractability
▪ Art =f(knowledge, experience, innate endowment, etc) – Apprenticeship
o Science: Apply scientific knowledge in mathematics, statistics, theory (economic, physical,
biological, etc)

[Link].3 ANOVA
• In practice/research, all we are trying to do is account for/explain the various sources of variation in the
phenomenon at hand
o ANOVA

• Hence, we include in the model all the variables we think are relevant for our purpose and dump/deposit the rest
into the “dustbin” called the “disturbance term”

4
E.g.
Y=f(x) ➔ Y = b0+ b1X1+ b2X2+ b3X3+ e [e=disturbance term]
o Unobservable
o Captures
▪ Errors in measurement
▪ Inefficiency
• Inability to make the right decisions/choices
o Probably due to lack of knowledge/skills/information/cognition/etc
▪ Voting??
▪ Input combination
▪ Uncertainty/Risk:
• Random (statistical) shocks (aka “white noise”), e.g., rainfed agriculture climate
related shocks
• White noise = variations +9in your data that cannot be explained by the regression
model

o Model without a disturbance term ➔ “deterministic” model/Mathematical model, e.g., 𝐸 = 𝑚𝑐 2

o Model with a disturbance term ➔ “stochastic” model/Statistical model

E.g.,
Y=f(x) ➔ Y = b0+ b1X1+ b2X2+ b3X3+ e [e=disturbance term]

5
[Link].4 Model syntax
[Link].4.1 Background/Theory
• Is a set of rules that must be followed when developing &/or using a model/or when simulating something
• Therefore, before any statistical analysis of economic data can be done, one needs a clear mathematical
formulation of the underlying theory that anchors your analysis [quantitatives = 1st term MSc]
o The choice of the mathematical formula depends on the phenomenon of interest:

EXAMPLES
(a) Biology
• Population growth model as exponential growth model: 𝑃𝑡 = 𝑃0 𝑒 𝑟𝑡
• Sigmoid/Logistic/Exponential curve (See Topic 4)

(b) Physics
▪ Newton’s gravity model

6
 • Used in Economics to estimate trade flows between 2 trading partners (See Topic 10)
(c) Chemistry
• Law of diffusion?
o Innovation diffusion theory

• Laws of thermodynamics and entropy, steady state and Walra’s general equilibrium (e.g., see
[Link]

• Von Liebig principle = Law of the minimum

(d) Sociology
• study of social relationships and systems
o Very many models (e.g., social mobility, migration, demography, etc)

(e) Psychology
• study of the human mind and its functions, especially those affecting behaviour in a given context
o Very many models (e.g., decision, choice, behavior theories)

(f) Economics – both micro & macroeconomics

• All what you’ve learned todate

• You need to understand the rules for you to develop a good model

Example: Equation of a Straight Line

𝑦 = 𝑚𝑥 + 𝑐
o LHS = dependent/outcome variable
o RHS => 𝑥 = independent/influencing/effects variable; 𝑐 = general mean/intercept; 𝑚 = slope

Graphically:

7
NOTE:
• You CANNOT interchange 𝒚 and 𝒙
• If we do, the results will be meaningless [bad semantics] [see below]
• REM: You will communicate your results to (i) peers, (ii) practitioners – policy makers & industry
o “Cooked” the results → cook the data
▪ Ethics: Values? What is your belief/value system?

[Link].4.2 What rules?

[Link].4.2.1 Mathematical rules

• Very many → from nursery school to present! RECALL!!

[Link].4.2.2 Statistical rules

[Link].[Link] Mathematical rules [see above]
[Link].[Link] Statistical assumptions concerning the generation of sample data from a population
E.g.
(a) Normality
RECALL: “Bell-shaped curve” => Central Limit Theorem [CLT]
-Matter/numbers/pop??
o When 𝑛 is large enough, the sampling distribution is approximately normal [bell-shaped curve]
because of the CLT

8
(b) Linearity vs non-linearity
• Linear → highest index on both variables & parameters =1
• Non-linear → either
o highest index on some variables >1, or
o the dependent variable, 𝑌, might take a restricted set of values or non-negative real numbers or
non-negative integers
• E.g.,
𝑌 = 0,1, 2,…, etc [rare events/count data]
𝑌 = 0 or 1
𝑌 = Category: 1=Car, 2=Air, 3=Train, 4=Bus

o Choice of parameter estimator – See later

▪ Ordinary least squares (OLS) technique + its variants
▪ Maximum likelihood (ML) technique
▪ Method of moments

(c) Independence
• Does the value of one variable influence or affect the value of other variable(s)?
• Lack of independence means linearity/singularity; you can represent one variable with the other

(i) Among regressors, 𝑿

• Difficult to compute the determinant of the 𝑿 matrix, i.e., |𝑿𝑇 𝑿|, that is needed in the computation of
unknown parameters – see later

(ii) Between regressors & the error term

• Problem of endogeneity ➔ results in biased parameter estimates – see later

(iii) Between the error terms in an equation system

𝐸(𝑢𝑖 , 𝑣𝑖 ) ≠ 0 ; 𝑐𝑜𝑣(𝑢𝑖 , 𝑣𝑖 ) ≠ 0➔ contemporaneous correlation in multiple equation models such as

simultaneous equations systems (SES) and seemingly unrelated regressions (SUR)

(d) Nature of the variance

(i) Size of the variance – large or small?
• Large variances ➔ high variability in the variable in equation

9
 o Indicative of problems with data: study design; collection, cleaning, transformation, etc
• Large variances are not good for hypothesis testing [Inference] – Why?

(iii) Approximation of “truth”

• Difficult → tools + expensive [time & money]
• Necessary
o Stochasticity ➔ range of values/“beliefs” about the “truth”
▪ Confidence about our “belief”

AVA_K = b0 + b1Uhuru+b2Rain_k+b3Lngdpp_k+e
• Our preoccupation is to the estimate the unknown parameters → betas

• After estimation: beta-hat; 𝛽̂𝑖

𝛽̂1 − 𝛽1
𝑡=
𝑆𝐸
Hypotheses testing [Inference]
𝐻0: 𝛽𝑖=0
𝐻1: 𝛽𝑖≠0

𝜎2
𝑆𝐸 = √
𝑛
̂
𝛽1 − 𝛽1
𝑡=
2
√𝜎
𝑛

. reg ava_k [Link] rain_k lngdpp_k

Source SS df MS Number of obs = 33

F(3, 29) = 6.86
Model .018275303 3 .006091768 Prob > F = 0.0012
Residual .025753668 29 .000888058 R-squared = 0.4151
Adj R-squared = 0.3546
Total .044028971 32 .001375905 Root MSE = .0298

ava_k Coef. Std. Err. t P>|t| [95% Conf. Interval]

[Link] .0213714 .0248193 0.86 0.396 -.0293898 .0721326

rain_k .0000454 .0000456 1.00 0.328 -.0000479 .0001387
lngdpp_k -.2968895 .1021878 -2.91 0.007 -.505887 -.087892
_cons 2.328791 .7177023 3.24 0.003 .8609248 3.796657

o Recall
▪ Type II error → You fail to reject 𝑯𝟎 when you shouldn’t [False negative]
▪ Type I error → You reject 𝑯𝟎 when you shouldn’t [False positive]

10
 • Small variances are better but may be unachievable – cost [time & money] + nature of the underlying
behavior

(ii) Is the variance constant or varying? I.e., homoscedastic vs heteroscedastic?

[Link].[Link] Rules of probability in statistics [statistical models]

• Basic probability rules

(a) Probability values range between 0 & 1

(b) Sum of all probabilities = 1
E.g., if probability of “success” = 𝑝 & that of “failure” = 1 − 𝑝, then, the sum of probability of “success” &
“failure” = 𝑝 + (1 − 𝑝) = 1
• This is related to the complement rule that the sum of the parts makes a whole
o If 𝑞 = 1 − 𝑝, then 𝑝 + 𝑞 = 1 and 𝑝 = 1 − 𝑞 while 𝑞 = 1 − 𝑝

(c) Addition rule

• The probability that one or 2 events occur is given as

(i) For mutually exclusive events: 𝑃(𝐴 or 𝐵) = 𝑃(𝐴) + 𝑃(𝐵)

(ii) For not mutually exclusive events: 𝑃(𝐴 or 𝐵) = 𝑃(𝐴) + 𝑃(𝐵) − 𝑃(𝐴 and 𝐵)

(d) Multiplicative rule

• The probability that 2 events occur together is given by

(i) For independent events: 𝑃(𝐴 and 𝐵) = 𝑃(𝐴) ∗ 𝑃(𝐵)
(ii) For non-independent events: 𝑃(𝐴 and 𝐵) = 𝑃(𝐴) ∗ 𝑃(𝐵|𝐴)

(e) Conditional probability ➔ the probability of an event happening given that another event has already
happened is given as:
𝑃(𝐴|𝐵) = 𝑃(𝐴 𝑎𝑛𝑑 𝐵)/𝑃(𝐵)

11
(f) Bayesian theory [Bayes’ theorem]
• Describes the probability of an event, based on prior knowledge of conditions that might be related to the
event

[Link].4.2.3 Falsification of theory [See Karl Popper (1963): Science as Falsification]

• For a theory to be considered scientific, it must be able to be tested and conceivably proven false
RECALL:
• A model is a theory spelled out in fully testable terms

Karl Popper (1963, p. ):

3. Every "good" scientific theory is a prohibition: it forbids certain things to happen. The more a theory
forbids, the better it is.

4. A theory which is not refutable by any conceivable event is non-scientific. Irrefutability is not a virtue of a
theory (as people often think) but a vice.

5. Every genuine test of a theory is an attempt to falsify it, or to refute it. Testability is falsifiability
6. Confirming evidence should not count except when it is the result of a genuine test of the theory

IN SUMMARY:
• The criterion of the scientific status of a theory is its falsifiability, or refutability, or testability

WATCH:
Schools of Economic Thought:
[Link]
[Link]

[Link].4.2.4 Rules in Computer algorithms

(a) STATA
(b) R
(c) Eviews
(d) Operations Research
-GAMS?
-LINDO?
(e) Others – Big Data Analytics & Machine Learning
-C++?
-Python?

12
[Link].5 Model Semantics
• Refers to the “meaning”/relevance of the model (results)
o As we said: “garbage in garbage out” → you might fit a model that is meaningless

• Such meaning/relevance is based on:

(i) the correct application of the model following appropriate rules, i.e., model syntax
• Knowledge of model specification
o Requires a clear understanding of the theory [economic, physical, biological, sociology, game,
psychology, etc] underpinning the behavior being modeled
▪ Literature review → in which realm of theory does your study belong
• Chapter 2 of your dissertation/report → “Review of theoretical literature”
(ii) knowledge of computer applications used to estimate the model
• Choice of appropriate computer application ➔ packaged software
o Use the latest version of the syntax
• Knowledge of computer coding ➔ unpackaged software
(iii) knowledge on how to interpret the results
• What story are you trying to tell your audience?
• Which audience?
(i) Peers in research & academia
(ii) Practitioners
a. Policy makers
b. Industry players
o Communication skills
▪ Creative both in writing & oral presentation
o Don’t be afraid to make mistakes – are part of learning
▪ Patience?
▪ Dexterity?
▪ Resilience?
o You need to “bring your head along”!

13
1.1.3 Types of models
• Economic model is a mathematical expression with a set of assumptions that approximately describes the
behavior of an economy or sector of an economy. The model includes all relevant variables without a “disturbance
term”
• An econometric model has the following features:
o Set of behavioral equations [mathematical expressions] derived from the economic
model/phenomena – the equations have (a) observed variables & (b) “disturbances” – all variables
considered irrelevant for the purpose of the model + any unforeseen events [uncertainty]
o Statement of whether there are errors in observation in the observed variables
o Specification of probability distribution of “disturbances” (& errors of measurement)
• With these specifications, we can test the validity of the economic model & use it to make
forecasts or use it in policy analysis

• Mathematical model = estimates/depicts an exact or “deterministic” relationship – no error/disturbance term

o E.g., in physics: E=mc2
o Mathematical models are solved using iteration ➔ numerical methods : Y=f(x) ➔ Y = b0+ b1X1+
b2X2: E=mc2
o y=Y Exact
o Area of circle = pie r^2

• Statistical model = adds a stochastic component into a deterministic relationship to change it from an exact
statement to a probabilistic description about expected outcomes – more useful to the econometrician

Approximate measure: 𝑦 ≈Y = b0+ b1X1+ b2X2+e [(i) uncertainty of outcome + (ii) errors in measurement of either Y or
X or both]

Demand for soda = f(-price, income, educ, age, weather, location, error term) – philosophy of parsimony – too big
or too small [1X →5-10 observation]

Amenable to policy
Art → experience
Science

1.1.4 Data
• Facts and statistics [observations; numbers] collected together for reference or analysis
• Depend on how they are collected
• 3 main types:
o Cross-sectional data – collected at a given point in time over space [spatial variability] or different
individuals [one-off survey]
▪ Time is not an important attribute of this type of dataset
o Time series data – collected at different points in time – mainly at fixed intervals – daily, weekly,
monthly, quarterly, annually, etc:
▪ Repeated over time but not on the same subject = Pooled data = combination of cross-
sectional and time series data e.g. cattle market prices [data for a particular animal are
cross-sectional but data for the market are time series]
• Your analysis account for
1. Cross-sectionness of the dataset [wrt the animal]
2. Repeatedness of the dataset [wrt the market]
▪ Repeated over time on the same subject [economic agent; DMU] = longitudinal or panel
data or micropanel data

14
 • Sentinels = group of subjects, countries, sectors
• Waves
• Your analysis account for
1. Repeatedness of the dataset [wrt the market]

[Link] Variables
• Is a measurable characteristic/attribute that varies
• 2 types:
o Quantitative variables – for quantity measurements (weight, income, output, prices, etc)
▪ Of 2 types:
i. Continuous or ratio variables = measurements of continuous or non-finite values
[from -∞ to +∞] e.g., 50.4; 50.9;
ii. Discrete or integer variables = Counts of individual items or values, e.g., 0, 1, 2,
3, etc
▪
o Qualitative or categorical variables – for quality measurements (color, taste, preference, choice,
presence/absence, etc)
▪ Of 3 types:
i. Binary [=dummy] variables = two levels denoting presence or absence [0, 1];
Coding: 1=Yes; 0=No ➔ convention/tradition
Coding: 0=Yes; 1=No
Coding: 1=Male; 0=Female
Coding: 0=Male; 1=Female
BETA COEFFICIENTS:
• Coefficient reported is one with 1=Yes; 0=No is taken a
baseline/reference value
• The interpretation of the beta coefficient is ALWAYS relative to the
reference/baseline value/case
ii. Nominal variables = groups with neither rank nor order between them
e.g., Likert scale 1=Strongly agree; 2=Agree; 3=Neutral; 4=Disagree; 5=Strongly
disagree
• Common in preference questions/research
iii. Ordinal variables = groups that are ranked in a specific order
e.g., on a rating scale
Taste: Scale of 1 to 10 ➔ 1 = least preferred; 10=most preferred

Model choice:
Y=f(x) ➔ Y = b0+ b1X1+ b2X2+e
• Y = dependent variable
• X = independent variable
o They can take different values – either qualitative or quantitative

A BIGGG NOTE
• The type of model chosen will depend on the nature of the dependent variables – is the dependent variable
qualitative or quantitative?
o Quantitative dependent variable ➔ regression model [OLS] → Chapters 2 & 3, 7 [with time as an
element]
o Qualitative dependent variable ➔ Chapter 4

15
 ▪ Binary [choice] model → Y takes only 2 values, 0 or 1
• LPM
• Logit
• Probit
▪ Multiple or discrete [choice] model → Y takes > 2 values
• Multinomial logit/probit – assumes independent choices
o Unordered
o Ordered
• Multivariate probit – assumes dependent choices
▪ Limited dependent variable model → Y takes 0 and non-zero values
• Y is a count data ➔ Poisson
• Y is censored between limits ➔ Tobit
• Y is truncated ➔ Truncated regression

▪ Nature of decision/behavior
• Sequential – sequential models
• Stepwise – Heckman

1.2 Purpose of Econometrics

• 3 main purposes of econometrics:
o Structural analysis
o Forecasting/prediction
o Policy evaluation

1. Structural analysis
• Involves use of an estimated econometric model for quantitative measurement of economic relationships
• Also facilitates testing of [economic] theory: comparison of rival theories of the same phenomenon
o A major result is feedback of the theory – either refuting or confirming the theory ➔ falsification
of a theory (See Karl Popper’s theory of falsification)
o Set our null hypothesis to be untrue ➔ based on [economic] theory
Steps (Deductive reasoning)
1. Set your null hypothesis based on economic/socio/psychology theory;
intuition/experience/hunch
2. Collect data
3. Analyze the data to get results
4. Use the results to test [confirm or refute] your hypothesis
5. Use the results for policy prescription

2. Forecasting/Prediction
• Is the use of an estimated econometric model to predict quantitative values of certain variables outside the
sample data actually observed
o In-sample prediction
▪ Involves using the same dataset that was used to train a model to evaluate its performance.
Essentially, you're testing the model on the dataset it already knows
▪ This can reveal how well the model "fits" the training data.

16
 o Out-of-sample prediction
▪ Involves evaluating a model's performance on data that were not used during the
training process. These data are "new" to the model.
▪ It provides a more realistic assessment of how well the model will generalize to unseen data.

In both cases:
Step 1: Get sample →gather data →analysis → estimated model: Yhat = b0_hat+b1_hatX1+b2_hatX2

Step 2: Apply the estimated model -- Yhat = b0_hat+b1_hatX1+b2_hatX2

o We use the numerical estimates of coefficients in order to forecast future values of the economic
phenomenon under study

• Prediction/forecasting is commonly used in financial/economic analyses – mainly using time series data

3. Policy evaluation
• Researcher: Data collection -→ analyze → report/policy recommendations
• Your results are an input into policy & practice -→ policy maker [monetary]; practitioners [NGOs, development
partners, etc]

• Involves the use of an estimated econometric model to choose between alternative policies
o Policy maker is resource-constrained
o Maximize objective function: Policy choices aimed to maximize social welfare
s.t. resource constraints
o Evidence-based policy making
• E.g. You can introduce an explicit objective function to be maximized/minimized by choice of policy
• OR: You can simulate alternative policy scenarios & make conditional forecasts of the future values of relevant
variables under each alternative [See chapter 9 – CGE modeling]

SAM -→CGE [=GAMS] ➔outputi

1.3 Methodology of Econometrics

• Basically, econometrics involves the following:
1. Statement of theory or hypothesis [=a proposition made as a basis for reasoning, without any
assumption of its truth. It is made on the basis of limited evidence to serve as a starting point for
further investigation] → falsification [confirm or refute your proposition/supposition]
2. Specification of the mathematical model of the theory: Y=f(x)
3. Specification of the statistical, or econometric, model: Y=f(x, e)
4. Obtaining data – Research methods
5. Estimation of parameters of the chosen model
6. Hypothesis tests [or statistical inference] → confirm or refute your proposition/supposition]
7. Forecasting or prediction
8. Use of the model for control or policy purposes

17
SUMMARY

Example [From Gujarati (2003)]

1. Statement of theory or hypothesis
• John Maynard Keynes stated that: men [women] are disposed, as a rule and on average, to increase their
consumption as their income increases, but not as much as the increase in their income
• This suggests that the marginal propensity to consume (MPC) or the rate of change of consumption for a unit
change in income is greater than zero but less than unity: 0<MPC<1.
2. Specification of the Mathematical Model of Consumption
• 0<MPC<1 can be stated mathematically as:
𝑌 = 𝛽1 + 𝛽2 𝑋 0 < 𝛽2 < 1 (1)
• where Y = consumption expenditure, X = income & 𝛽1 and 𝛽2 are unknown parameters to be estimated
• Eqn 1 is a single eqn deterministic model; we can have multiple eqn models [see Topic 5 in the Course
outline]
• Graphically, the mathematical model looks like:

Figure 1a. Relationship between income & consumption expenditure in a mathematical model

18
3. Specification of the Econometric Model of Consumption
• We add a disturbance or error term or random (stochastic) variable, 𝒖, to eqn 1 to convert it from
mathematical to a statistical or econometric [or statistical] model:
𝑌 = 𝛽1 + 𝛽2 𝑋 + 𝒖 0 < 𝛽2 < 1 (2)

• Graphically, the econometric model looks like [which allows for an inexact relationship between Y and X]:
o Inexactness arises from (i) errors in measurement of either Y or X, (ii) omitted variables

Figure 1b. Relationship between income & consumption expenditure in an econometric model

4. Obtaining data – REFER TO YOUR RESEARCH METHODS CLASS

• Check Gujarati (2003) for the rest of the steps

1.4 Types of Econometrics

• Econometrics can be divided into 2 main branches – theoretical & applied

Figure 2: Types of econometrics

• Theoretical econometrics – is concerned with the development of appropriate methods for measuring economic
relationships specified by econometric models

19
 o It states assumptions of the method, its properties & consequences for violating assumptions – e.g.,
in the least squares method

• Applied econometrics – uses tools of theoretical econometrics to study economic phenomena, e.g., production
function, investment function, demand & supply function, portfolio theory, etc

• Both theoretical & applied econometrics are based on either classical [theory of marginalism ➔ Alfred
Marshall/Marshallian] or Bayesian [probability theory] tradition
o We use the classical tradition

20