Math, Stats, and

Mathstats Review
ECONOMETRICS (ECON 360)
BEN VAN KAMMEN, PHD
Outline
These preliminaries serve to signal to students what tools they need to know to succeed in
ECON 360 and refresh their familiarity with these tools. These are things you would be expected
to learn in introductory college courses in:
◦ Algebra,
◦ Calculus,
◦ Probability and Statistics.

They are also addressed in the appendices (particularly A, B, C) of the Wooldridge text (which is
required for this class).
The summation operator
Notation short hand for representing the sum of a set of observations.
◦ Consists of an index (usually i) that uniquely identifies the observations,
◦ the sum operator, Σ,
◦ and the terms that must be added for all together for a given number of observations (n).

I.e.,
𝑛𝑛

� 𝑥𝑥𝑖𝑖 = 𝑥𝑥1 + 𝑥𝑥2 +. . . 𝑥𝑥𝑛𝑛 .

𝑖𝑖=1
The summation operator (continued)
The sum operator has some convenient properties:
1. Summing constants (anything not indexed with an i) does not really require a sum operator.
𝑛𝑛

� 𝑐𝑐 = 𝑐𝑐 + 𝑐𝑐+. . . +𝑐𝑐 = 𝑐𝑐 1 + 1+. . . 1 = 𝑐𝑐𝑐𝑐.

𝑖𝑖=1
The summation operator (continued)
2. Constants can be “pulled through” or factored out of a term using a sum operator.
𝑛𝑛 𝑛𝑛

� 𝑐𝑐𝑥𝑥𝑖𝑖 = 𝑐𝑐 � 𝑥𝑥𝑖𝑖 .
𝑖𝑖=1 𝑖𝑖=1
3. Sums are commutative.
𝑛𝑛 𝑛𝑛 𝑛𝑛

�(𝑎𝑎𝑥𝑥𝑖𝑖 + 𝑏𝑏𝑦𝑦𝑖𝑖 ) = 𝑎𝑎 � 𝑥𝑥𝑖𝑖 + 𝑏𝑏 � 𝑦𝑦𝑖𝑖 .

𝑖𝑖=1 𝑖𝑖=1 𝑖𝑖=1
The summation operator (concluded)
4. The sum of a fraction does not equal the fraction of the sums.

𝑛𝑛
𝑥𝑥𝑖𝑖 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖
� ≠ 𝑛𝑛 .
𝑦𝑦𝑖𝑖 ∑𝑖𝑖=1 𝑦𝑦𝑖𝑖
𝑖𝑖=1
There are other clever “tricks” that involve simplifying the terms inside a sum operator that are
useful and will be demonstrated later.
Functions
Function: a rule that associates a value of (variable) x with exactly one value of (variable) y.
When y is a function of x, write 𝑦𝑦 = 𝑓𝑓(𝑥𝑥).
Economics applications:
◦ demand and supply functions that associate a price with a quantity demanded or supplied,
◦ production functions (associate quantities of inputs with quantities of outputs),
◦ cost functions (associate quantities of outputs with costs incurred),
◦ profit functions (quantity and profit),
◦ utility functions (quantity and utility).
Functions (continued)
Functions can involve:
◦ the basic operations: “+”, “-”, “×”,“÷”,
◦ exponents and roots,
◦ e.g., 𝑓𝑓(𝑥𝑥) = 𝑥𝑥2. The superscript “2” (the exponent) indicates the power to which x is raised.
1
◦ e.g., 𝑓𝑓(𝑥𝑥) = 𝑥𝑥. This is the square root of x. Can be expressed as : 𝑓𝑓 𝑥𝑥 = 𝑥𝑥 = 𝑥𝑥 2 .
◦ constants,
◦ e.g., 𝑓𝑓(𝑥𝑥) = 𝑥𝑥 + 5. In this example 5 is a constant.
◦ and coefficients.
◦ e.g., 𝑓𝑓(𝑥𝑥) = 10𝑥𝑥. In this example 10 is the coefficient.
Kinds of functions
Only variables, constants, and coefficients (no exponents) → linear function.
◦ Form is: 𝑦𝑦 = 𝑓𝑓(𝑥𝑥) = 𝑚𝑚𝑚𝑚 + 𝑏𝑏, where m is a coefficient and b is a constant.
Form: 𝑦𝑦 = 𝑓𝑓(𝑥𝑥) = 𝑎𝑎𝑎𝑎2 + 𝑏𝑏𝑏𝑏 + 𝑐𝑐, is a quadratic function.
Power function: 𝑦𝑦 = 𝑓𝑓(𝑥𝑥) = 𝑎𝑎𝑥𝑥 𝑏𝑏 .
Exponential function: 𝑦𝑦 = 𝑓𝑓 𝑥𝑥 = 𝑎𝑎𝑏𝑏 𝑥𝑥 .
◦ b is the base for the exponent x.
Logarithmic function: 𝑦𝑦 = 𝑓𝑓 𝑥𝑥 = log 𝑏𝑏 𝑥𝑥.
◦ Again b is the base.
◦ y is the power to which you must raise x to get b.
◦ The base in many applications is “Euler’s number”: 𝑒𝑒 ≈ 2.718.
Graphing linear functions
2 known points or the slope-intercept form.
◦ Since these functions are always a straight line, you just need to know two points that the line goes
through; you can then connect them with a line.
◦ If you know that the coefficient is the slope and the constant is the y intercept, you can graph the line
using that information.
A linear function without slope
A linear function with slope only

Note that the slope here is positive (a>0); slope can be negative, too.
A linear function with slope and constant

This is a case where the slope of the line is negative (a<0).

Slope of a linear function
The slope of a linear function is the same along its entire length.
◦ The slope is constant.

The function, 𝑓𝑓(𝑥𝑥) = 4𝑥𝑥, has a slope of 4—which is a constant.

The simplicity of constant slope is what makes linear functions so useful for our purposes in
economics classes.
As you will see, the slope of a function has very powerful intuitive implications for econometric
analysis.
◦ More on the other functional forms as necessary, later in the course.
Differential calculus
Non linear functions have variable slope.
To calculate slope, differentiate the function instead of differencing (“rise over run”).
◦ Result is called the first derivative.

Example: 𝑦𝑦 = 𝑓𝑓 𝑥𝑥 = −2𝑥𝑥 2 + 4𝑥𝑥.

◦ Rise over run at 𝑥𝑥 = 3 is:
𝑓𝑓 4 − 𝑓𝑓 3
= −16 − −6 = −10.
4−3
◦ Gives you the slope of the secant line connecting the 2 points.
Differential calculus (continued)
Differential calculus (continued)
Differencing in the limit as ∆𝑥𝑥 → 0.
◦ Moving the 2 points infinitely close together.
◦ Replace ∆𝑥𝑥 with notation 𝜀𝜀.
◦ Again beginning at 3:
∆𝑓𝑓 𝑥𝑥 −2 9 + 6𝜀𝜀 + 𝜀𝜀 2 + 4 3 + 𝜀𝜀 − (−6)
𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = = = −2𝜀𝜀 − 8
𝜀𝜀 𝜀𝜀
Taking the limit as change goes to zero simply gives you,
𝛿𝛿𝑦𝑦
= −8.
𝛿𝛿𝑥𝑥
First derivative: the slope of the line
tangent to the function
Differentiation in practice
“Cookbook” procedure for differentiation.
◦ Multiply the coefficient on each term by the exponent on that term.
◦ Reduce the exponent on each term by one.
◦ For this example,
𝑓𝑓 𝑥𝑥 = −2𝑥𝑥 2 + 4𝑥𝑥,
the first derivative is,
𝛿𝛿𝑓𝑓(𝑥𝑥)
= −2 2 𝑥𝑥 2−1 + 4 1 𝑥𝑥 1−1 = −4𝑥𝑥 + 4.
𝛿𝛿𝑥𝑥
Partial differentiation
When functions have more than 1 argument, they have more than 1 derivative.
◦ Just have to specify the variable with which you are differentiating, e.g.,
𝑦𝑦 = 𝑓𝑓 𝑥𝑥, 𝑧𝑧 .
Notation:
𝜕𝜕𝜕𝜕
◦ , or
𝜕𝜕𝜕𝜕
◦ 𝑓𝑓𝑥𝑥 , or
◦ 𝑦𝑦𝑘𝑘 .

Economics examples:
◦ Marginal utility,
◦ Marginal product of labor.
Partial differentiation (continued)
Differentiate with respect to 1 independent variable at a time.
◦ Treat other arguments as constants.
◦ Extra easy when there are no interaction terms (in which 2 variables multiply together).

Example: 𝑄𝑄 𝐿𝐿, 𝐾𝐾 = 𝐴𝐴𝐾𝐾 𝑎𝑎 𝐿𝐿1−𝑎𝑎

𝜕𝜕𝑄𝑄
= 𝑎𝑎𝑎𝑎𝐾𝐾 𝑎𝑎−1 𝐿𝐿1−𝑎𝑎
𝜕𝜕𝐾𝐾
and
𝜕𝜕𝜕𝜕
= 1 − 𝑎𝑎 𝐴𝐴𝐾𝐾 𝑎𝑎 𝐿𝐿−𝑎𝑎 .
𝜕𝜕𝐿𝐿
Tabular descriptive statistics: cross
tabulation
A tabular summary of data for two variables.
Cross tabulation can be used when:
◦ one variable is nominal/ordinal and the other is ordinal/interval/ratio,
◦ both variables are ordinal/interval/ratio, or
◦ both variables are nominal/ordinal.
Cross tabulation (continued)
Cross tabulation (continued)
Recall what a frequency distribution for one variable does.
Cross tabulation as multiple frequency distributions:
◦ one for each value of a second variable.

More information by expressing frequencies relative to either the column or row totals.
◦ Analogous to the relative (or percent) frequency distribution.
Cross tabulation, column percentages
Cross tabulation, row percentages
Cross tabulation (concluded)
Column percentages:
◦ each column sums to 100%.
◦ Note that the rows do not sum to 100%.
◦ Vice versa for row percentages.

The two tables have slightly different interpretations.

◦ In the first one, we're comparing the distributions of workers' travel behavior based on how many kids
they have.
◦ In the second, we're comparing the distributions of their parenthood responsibilities based on how
often they travel.
Graphical descriptive statistics:
scatterplot
A graphical presentation of the relationship between two quantitative variables.
One variable is shown on the horizontal axis and the other variable is shown on the vertical axis.
The general pattern of the plotted points suggests the overall relationship between the
variables.
A trend line can be added as an approximation of the relationship.
Scatterplot (continued)
Scatterplot (continued)
Numerical multivariate descriptive
statistics
Measures of association between two variables
◦ quantify visual evidence of a relationship.

They are covariance (𝑠𝑠𝑥𝑥𝑥𝑥 ) and the correlation coefficient (𝑟𝑟𝑥𝑥𝑥𝑥 ).

◦ Note these are sample statistics that estimate the true underlying relationship in the population:
𝐸𝐸 𝑠𝑠𝑥𝑥𝑥𝑥 = 𝜎𝜎𝑥𝑥𝑥𝑥 and 𝐸𝐸 𝑟𝑟𝑥𝑥𝑥𝑥 = 𝜌𝜌𝑥𝑥𝑥𝑥 population covariance and correlation .
Measures of association (continued)
Positive values indicate a positive relationship, and negative values indicate a negative
relationship.
𝑛𝑛
1
𝑠𝑠𝑥𝑥𝑥𝑥 ≡ � 𝑥𝑥𝑖𝑖 − 𝑥𝑥̅ 𝑦𝑦𝑖𝑖 − 𝑦𝑦� and,
𝑛𝑛 − 1
𝑖𝑖=1
𝑠𝑠𝑥𝑥𝑥𝑥
𝑟𝑟𝑥𝑥𝑥𝑥 ≡ ; 𝑠𝑠𝑥𝑥 and 𝑠𝑠𝑦𝑦 are sample standard deviations of 𝑥𝑥 and 𝑦𝑦.
𝑠𝑠𝑥𝑥 𝑠𝑠𝑦𝑦
To demonstrate how these definitions imply positive and negative signs, consider a scatterplot
and the point, (𝑥𝑥,̅ 𝑦𝑦).
�
◦ Variables with a “bar” over them usually indicate the sample mean.
Measures of association (continued)

The sample means are the intersection of the red lines in the graph.
Measures of association (continued)
The point (𝑥𝑥,̅ 𝑦𝑦)
� divides the plot into four regions:
◦ The “north-east”: in which 𝑥𝑥𝑖𝑖 > 𝑥𝑥̅ and 𝑦𝑦𝑖𝑖 > 𝑦𝑦.
�
◦ The “south-east”: in which 𝑥𝑥𝑖𝑖 > 𝑥𝑥̅ and 𝑦𝑦𝑖𝑖 < 𝑦𝑦.
�
◦ The “south-west”: in which 𝑥𝑥𝑖𝑖 < 𝑥𝑥̅ and 𝑦𝑦𝑖𝑖 < 𝑦𝑦.
�
◦ The “north-west”: in which 𝑥𝑥𝑖𝑖 < 𝑥𝑥̅ and 𝑦𝑦𝑖𝑖 > 𝑦𝑦.
�
Measures of association (continued)
Consider the product, 𝑥𝑥𝑖𝑖 − 𝑥𝑥̅ 𝑦𝑦𝑖𝑖 − 𝑦𝑦� for an observation in each of the four regions.
◦ For number 1, it is a positive times a positive.
◦ For number 3, it is a negative times a negative.
◦ For the other two, it's a negative times a positive.

For the odd-numbered regions, the product will be positive, and for the even-numbered regions,
it will be negative.
Covariance
Recall the appearance of a positive relationship:
◦ it should have a lot of observations in the “north-east” and “south-west” parts of the scatterplot.
◦ Similarly a negative relationship should have a lot of points in the “north-west” and “south-east”.

If the data set has a lot of observations in the odd-numbered regions on the scatterplot, they
will count positively toward the covariance.
And observations in the even-numbered regions count negatively toward covariance.
If there are more of the former, covariance will be positive, and if there are more of the latter, it
will be negative.
◦ For the variables on the graph above, the covariance is roughly 15,234, a positive number—a positive
relationship between population and average wages across counties.
Correlation coefficient
Estimates the strength of the relationship
Divide covariance by the product of standard deviations.
◦ Standard deviation is always positive → preserves the sign of covariance.
◦ Standardizes the scale: measures with larger scales can be compared purely on the basis of how strong
the relationship is.

An “r” of 1 represents a perfect positive linear relationship, and an “r” of -1 is a perfect negative
linear relationship.
Zero is the dividing point between positive and negative association and represents zero
covariance and zero relationship between the two variables.
◦ The correlation coefficient for the population and wage (above) is 0.4583—a modestly strong positive
relationship.
Random variables
Numerical descriptions for the outcome of a probabilistic experiment.
If x is a random variable (r.v.), it takes on different values with a specific probability distribution.
◦ Discrete: a finite number of values or an infinite series of values.
◦ Continuous: can assume any (infinitely many) value on one or more interval(s).

Probabilities summarized by a probability function:

◦ Pr 𝑥𝑥 = 𝑥𝑥𝑖𝑖 = 𝑓𝑓 𝑥𝑥𝑖𝑖 ∈ [0,1]
◦ ∑𝑖𝑖∈𝐼𝐼 𝑓𝑓 𝑥𝑥𝑖𝑖 = 1
“That there, Clark, is an r.v.” – Cousin
Eddie . . . different kind of r.v. though
Describing a random variable
Mean or expected value or first moment of the distribution.
◦ A probability-weighted average of all the values the r.v. takes.

𝐸𝐸(𝑥𝑥) ≡ � 𝑥𝑥𝑖𝑖 𝑓𝑓 𝑥𝑥𝑖𝑖

𝑖𝑖∈𝐼𝐼
Variance or second moment.
◦ A probability weighted average of the squared deviations from the mean.
2
𝜎𝜎 2 = 𝑉𝑉𝑉𝑉𝑉𝑉(𝑥𝑥) ≡ � 𝑥𝑥𝑖𝑖 − 𝐸𝐸 𝑥𝑥 𝑓𝑓 𝑥𝑥𝑖𝑖
𝑖𝑖∈𝐼𝐼
Expected value
Properties.
◦ Expected value of a constant, c, is 𝐸𝐸 𝑐𝑐 = 𝑐𝑐.
◦ Constants “pull through” expectations: 𝐸𝐸 𝑎𝑎𝑎𝑎 + 𝑏𝑏 = 𝑎𝑎𝑎𝑎 𝑥𝑥 + 𝑏𝑏.
◦ Expected value is commutative: 𝐸𝐸 ∑𝑛𝑛𝑖𝑖=1 𝑎𝑎𝑖𝑖 𝑥𝑥𝑖𝑖 = ∑𝑛𝑛𝑖𝑖=1 𝑎𝑎𝑖𝑖 ∗ 𝐸𝐸 𝑥𝑥𝑖𝑖 .
Variance
Properties.
◦ The properties of expected value also apply because variance can be phrased as the expectation of:
2
𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥 = 𝐸𝐸 𝑥𝑥 − 𝐸𝐸 𝑥𝑥 = 𝐸𝐸 𝑥𝑥 2 − 𝐸𝐸 𝑥𝑥 2
.
◦ Additionally, iff Pr 𝑥𝑥 = 𝑐𝑐 = 1, the variance of x is zero.
◦ Constants “pull through”, a la,
2 2
𝑉𝑉𝑉𝑉𝑉𝑉 𝑎𝑎𝑎𝑎 + 𝑏𝑏 = 𝐸𝐸 𝑎𝑎𝑎𝑎 + 𝑏𝑏 − 𝐸𝐸 𝑎𝑎𝑎𝑎 + 𝑏𝑏 = 𝐸𝐸 𝑎𝑎 𝑥𝑥 − 𝐸𝐸 𝑥𝑥 , so
2 2
𝑉𝑉𝑉𝑉𝑉𝑉 𝑎𝑎𝑎𝑎 + 𝑏𝑏 = 𝐸𝐸 𝑎𝑎 𝑥𝑥 − 𝐸𝐸 𝑥𝑥 = 𝑎𝑎2 ∗ 𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥 .
Variance (continued)
More properties.
◦ Generalizing to the variance of a sum of random variables.
2
𝑉𝑉𝑉𝑉𝑉𝑉 𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏 = 𝐸𝐸 𝑎𝑎 𝑥𝑥 − 𝐸𝐸 𝑥𝑥 + 𝑏𝑏 𝑦𝑦 − 𝐸𝐸 𝑦𝑦
2 2
⇔ 𝑉𝑉𝑉𝑉𝑉𝑉 𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏 = 𝐸𝐸 𝑎𝑎2 𝑥𝑥 − 𝐸𝐸 𝑥𝑥 + 𝑏𝑏 2 𝑦𝑦 − 𝐸𝐸 𝑦𝑦 + 2𝑎𝑎𝑎𝑎 𝑥𝑥 − 𝐸𝐸 𝑥𝑥 𝑦𝑦 − 𝐸𝐸 𝑦𝑦
⇔ 𝑉𝑉𝑉𝑉𝑉𝑉 𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏 = 𝑎𝑎2 𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥 + 𝑏𝑏 2 𝑉𝑉𝑉𝑉𝑉𝑉 𝑦𝑦 + 2𝑎𝑎𝑎𝑎 ∗ 𝐶𝐶𝐶𝐶𝐶𝐶(𝑥𝑥, 𝑦𝑦)
◦ For pairwise uncorrelated variables (none is correlated with any other r.v.).

𝑉𝑉𝑉𝑉𝑉𝑉 � 𝑎𝑎𝑖𝑖 𝑥𝑥𝑖𝑖 = � 𝑎𝑎𝑖𝑖2 𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥𝑖𝑖 ,

𝑖𝑖∈𝐼𝐼 𝑖𝑖∈𝐼𝐼
because all the covariance terms are zero.
Conditional expectation
Same notion as expected value, but re-weighted with information generated by knowing the
value/range of a r.v., e.g.,
◦ 𝐸𝐸 𝑦𝑦 𝑥𝑥 = 6 ,
◦ 𝐸𝐸 𝑦𝑦 𝑥𝑥 ≥ 6 ,
◦ 𝐸𝐸 𝑦𝑦 𝑦𝑦 ≥ 6 .

Properties.
◦ Functions of r.v. behave like constants when you condition on the r.v.
𝐸𝐸 𝑐𝑐 𝑥𝑥 𝑥𝑥 = 𝑐𝑐 𝑥𝑥 , for any function, 𝑐𝑐
◦ Unconditional and conditional expectations are related by:
𝐸𝐸 𝑎𝑎 𝑥𝑥 𝑦𝑦 + 𝑏𝑏 𝑥𝑥 |𝑥𝑥 = 𝑎𝑎 𝑥𝑥 ∗ 𝐸𝐸 𝑦𝑦 𝑥𝑥 + 𝑏𝑏(𝑥𝑥)
◦ And
𝐸𝐸 𝑦𝑦 𝑥𝑥 = 𝐸𝐸 𝑦𝑦 iff 𝑥𝑥 and 𝑦𝑦 are independent.
Conditional expectation (continued)
Law of iterated expectations.
𝐸𝐸 𝐸𝐸 𝑦𝑦 𝑥𝑥 = 𝐸𝐸(𝑦𝑦)
◦ Take the expectation of the value in brackets across all different values of x and you arrive at the
unconditional expectation.

If x and y are independent, their correlation and covariance are zero.

Conclusion
Other things you learn in earlier classes (not specifically covered here) are used in ECON 360,
too.
◦ Distributions and their properties,
◦ Performing statistical inference,
◦ Optimization using differential calculus.

These topics are central to performing regression analysis and will be covered in more detail in
subsequent lectures.
The topics in this set of notes is so fundamental that I expect students to be familiar with them
prior to beginning to turning their attention to regression.
◦ Hopefully after this review, that is the case.