Chapter 3&4 Lecture Notes
Chapter 3&4 Lecture Notes
Chapter 3&4 Lecture Notes
3&4
Chapter 3
Objective Function – A function that describes firm behavior, it usually defines profit. It will
then be turned into a maximization or minimization problem.
Maximization Problem – Problem used to find optimal (maximum values of the objective
function, usually profit or output.
Minimization Problem – Problem used to find optimal (minimum) values of the objective
function, usually a cost function.
Choice Variable – Variables under the control of the firm, that determine the value of the
objective function. In this case it’s “q”, the individual firm gets to choose how much to produce.
Discrete Variable – Choice variables that can only take on limited values like integers: 1, 2, 3, 4
Continuous Variable – Choice variable that can take on any value between two points.
Unconstrained Optimization – Optimizing an objective function over all possible values of the choice
variable (q) without any conditions
Constrained Optimization – Optimizing an objective function over all possible values of choice variable
(q) where some other condition also holds. Usually a budget constraint.
Marginal Analysis – Using infinitesimally small changes to find a maximum/minimum in the objective
function.
Simple Optimization Example: Running a Bookstore (Unconstrained)
Every business owner runs into the same problem with a downward sloping demand curve
o If you lower your price, you sell more units, but you get less per price
o If you increase your price, you will get more per unit, but you sell fewer units.
What should you do?
o One solution is to calculate Total Revenue (TR = Price * Quantity)
An Easier Way?
Net Benefits
Most of these types of problems take this basic form NB = TB-TC
NB – Net Benefit, TB – Total Benefit, TC – Total Cost
You will pick some level of activity (A) (output q in the book example)
The activity A is usually assumed to have two properties.
o Always increasing in TB but at a decreasing rate.
o TC is always increasing, usually at an increasing rate.
Under these conditions, the value of A that yields the largest difference between TB and TC is
the one that will maximize Net Benefits (NB) aka Profit or Utility.
Marginal Benefit – Gives the measure of the change in benefits given a change in the activity (at a
particular point or between points).
Marginal Cost – Gives the measure of the change in costs given a change in the activity (at a particular
point or between points).
MB and MC define the slopes of the tangent lines to the TB and TC curves.
Total Costs – Sum of all costs of operation including fixed costs and variable costs, does NOT include
sunk costs.
Fixed Costs – Don’t vary. Costs that do not vary with output, must be paid
Sunk Costs – Costs that have already been paid and cannot be recovered. Should not be considered in
future decision making.
Constrained Optimization
The optimal solution, the one that maximizes utility, is one in which you consume in a way to
equalize the per dollar marginal benefits equal to one another.
When you have a budget, resource, or time constraint, you have to ration the scarce resources.
Where do these conditions come from?
Graphing beer consumed against total happiness (holding wing consumption constant)
Chapter 4
Parameters and Regression
A, b, and c are called parameters, and they need to be estimated.
Parameters are coefficients in an equation that determine the exact mathematical relation
among the variables in the equation.
We get an estimate of the parameters using a technique called regression analysis.
Regression analysis - technique of fitting a line through a series of points to minimize some
error function. The most common error function used is squared error.
Regression analysis is also used to find the values of parameters.
o Uses data on economic variables to determine a mathematical equation that describes
the relation between the economic variables.
o Objective of a regression analysis is to find.
Estimates of the true slope and intercept parameters
The straight line that best fits a scatter of data points.
A Regression equation shows the expected level of a dependent variable given levels of
explanatory variables.
Variables
Regression is used to determine the relationship between dependent variables and explanatory
variables.
Dependent Variables - variable that is said to “depend" or be “caused by" the explanatory
variables.
o Total cost in our example
Explanatory Variables - variables that “explain" or “cause" certain values of the dependent
variable.
o Quantity produced in our example.
Linear Model
In a linear model we ASSUME that the relationship is linear,
Y = a + bX
o a - “Intercept” parameter, where the line crosses the Y axis
o b - “Slope” parameter, it is the slope of the line
It is called a liner model because that is the generic equation of a line.
The True Model
For regression analysis to work, and to be meaningful, there has to be an actual or true
relationship between the variables.
X and Y need to actually be related to one another in some way.
The true relationship is unknown, that is the point of estimation, but it must exist or this whole
enterprise is a waste of time.
o Pickles and pregnancy
o Hot dogs and horseshoes
Example:
If that is the “true model,” the actual relationship, why don’t all of my observed points fall on the
line? The data probably looks something like this,
Ads are not the only thing that explain total dollar sales S!
o local income
o weather
o preferences
o word of mouth
We lump everything else into the random error or error term (e)
Random Error - all of the unobservable factors that influence S other than A
1. We could pick a line that minimizes the sum of all of the absolute values.
Time-Series data set – data collected over time for a specific firm.
Cross-Sectional data set – data collected from several different firms at a given time.
It can be shown that the line that minimizes the sum of the squared error is unique, which
is super convenient.
o
Estimates and Estimators
Statistical Significance
Statistically Significant - we have sufficient evidence (predicated on our assumptions being true) that the
TRUE value of the relationship between taxes and population (b) is not zero.
b v.s. bˆ - One is the true value we do not observe, the other is an estimate based o ff of a sample.
We are taking a sample and making an inference about all possible cases using that sample.
A parameter is said to be statistically significant if there is sufficient evidence that the: true value
of the parameter does not equal zero.
Hypothesis Testing
Hypothesis Test - A technique where we estimate the probability of an outcome under some
hypothesized value.
Assume for a moment that there is no true relationship between taxes levied and population.
A typical hypothesis test calculates the probability that the sample of observations we pulled
would find a value of 209 given that there is no actual relationship between taxes and population.
There is no relationship between a person’s biological sex and their eye color.
However we could still test this. There is a true proportion, we could find it by grabbing all the
people with brown eyes and counting, but that is impossible.
We could walk around and collect observations (take samples of the population) on eye color and
sex.
Each day we collect a sample, there will be a di fferent proportion of male/female that have brown
eyes, just due to randomness.
Our estimate is the proportion of male/female with brown eyes.
Distribution of bˆ
Relative frequency distribution - The distribution (and relative frequency) of values b̂ can take
because observations on Y and X come from a random sample.
We can treat each day as a sample, each sample has its own estimate, and we could do this a lot.
We could save each one of these estimates, and before long we have a distribution of the
estimates.
The t-statistic
The larger the t-stat, the more confident we are in rejecting the null hypothesis. We usually pick some
critical value of t(more on this in a minute), prior to the investigation, that would cause us to reject
the null or hypothesized value. This value is typically anywhere from 1.96 to 2.00 in absolute value.
For simplicity assume Sbˆ = .1 do we reject a bh = .5?
0.2 is a far cry from 1.96, so we would fail to reject this null.
This makes sense because the hypothesized value is the true value.
Errors
Type I Error - finding a parameter estimate to be significant when it is not.
o We find a relationship between variables that does not really exist. We reject the null
hypothesis (bh = 0) for pop munic when we shouldn’t.
Type II Error - We fail to reject a null that should have been rejected.
o We find no relationship between variables that does indeed really exist.
Level of Significance/Confidence
Level of Significance - How willing are we to commit a Type I error (reject a true null)?
Just due to randomness in samples it is possible that we collect a sample that would cause us to
reject a true null of .5.
If we hypothetically grabbed 100 samples what percentage of those are we willing to reject a true
null?
The standard is 5% we set the critical t-values such that we would incorrectly reject a true null.
If the t-statistic value is greater than the t-critical , meaning that it is beyond it on the x-axis (a
blue x), then the null hypothesis is rejected and the alternate hypothesis is accepted. However, if
the t-statistic had been less than the t-critical value (a red x), the null hypothesis would have been
retained.
Level of Confidence - is (1 - Level of significance), if the Level of significance = 5% then we are 95%
confident in our decision to accept/reject the null.
The p-value
1. It is the probability of a Type I error (rejecting a true null) given that the null.
a. (bh = 0) is true.
b. The exact level of significance for a t-ratio associated with a parameter estimate.
2. it is the probability of finding the bˆ that we found, or something more extreme, given
a. that the true b = 0.
Yˆ
Yˆ = aˆ + bˆX
Coefficient of determination – Measures the fraction of the total variation in the dependent variable that
is explained by the regression equation.
The value of R^2 can rang from 0 to 1.
If the value R^2 is high, there is high correlation between the dependent variable and
independent variables and vice versa.
F-Statistic – Provides a measure of the ratio of explained variation (in the dependent variable) to
unexplained variation.
To test whether the overall regression equation is significant, this statistic is compared
with the critical F value obtained from an F-table.
K-1 and n-k degrees of freedom.
If the F-Stat exceeds the critical F-value, the regression equation is statistically
significant.
Multiple Regression
This nonlinear function form is used to estimate demand functions and to estimate production functions,
which is particularly useful because the parameters “b” and “c” are elasticities.
Using this form of nonlinear regression, the elasticities are estimated directly: The parameter estimates
associated with each explanatory variable are elasticities. (The parameter a, however, is not an elasticity).
To estimate the parameters of this nonlinear equation, it must be transformed into a linear form. This is
accomplished by taking natural logarithms of both sides of the equation. Taking the logarithm of the
function results in: