Page |1

Financial Econometrics
ASSIGNMENT: Functional Forms of Regression Models
By - Sagnik Monga (18363)
INDEX:
I. ABSTRACT
II. AIM
III. THE STUDY
a. LINEAR REGRESSION EQUATIONS
b. LOGARITHMIC MODELS
i. LINEAR-LOG MODEL
ii. LOG-LINEAR MODEL
iii. LOG-LOG MODEL
c. RECIPROCAL MODEL
d. POLYNOMIAL MODEL
IV. CONCLUSION
V. SCOPE FOR FUTURE STUDY
VI. SOURCES
ABSTRACT
Nonlinear regression analysis is a very popular technique in mathematical and social
sciences as well as in engineering. Ordinary Least Square method is the most widely used
approach to parameters estimation. Under the normality assumption of errors, the least
squares estimates equal the maximum likelihood estimates. In many cases, the dependent
variable in a regression model can be influenced by both quantitative variables and
qualitative factors. Functional forms come into play the magnitude of difference in the
dependent and independent variables is too huge to properly study and make out
significant deductions. In some cases, linearity alone is not sufficient to build a robust
model and analyse the error term to its full extent. Thus, log/exponential/reciprocal are
some of the ways that bring down the value of difference in the variables and make the
study more feasible. But the study is not just limited to these three forms, there is a whole
universe which can be explored combing the effect of multiple variables and multiple
functional forms.
Key Terms: Slope, Elasticity, Linearity, Over-fitting, Under-fitting
AIM: The aim of this assignment is to study the functional forms of regression models,
differentiating the logarithmic form with linear, polynomial, reciprocal forms, and stating the
practical examples and situations of using each different model.

Financial Econometrics | Functional Forms

 Page |2

THE STUDY
No regression model is perfect. The error term contains the influence of any factors
(variables) that affect your dependent variable and aren’t captured by your independent
variable(s). The characteristics of the error term are probably of greater importance than the
explained part. You need several assumptions about the error term to prove that the OLS
results are precise. The assumption that the error term is normally distributed isn’t required
for performing OLS estimation, but it is necessary when you want to produce confidence
intervals and/or perform hypothesis tests with your OLS estimates.
Linear Regression Equations
A linear regression model follows a very basic form. In statistics, a regression model is linear
when all terms in the model are one of the following:
• The constant (not always required)
• A “linear” parameter multiplied by an independent variable (IV)
Then, you build the equation by only adding the terms together. These rules limit the form
to just one type:
Dependent variable = constant + parameter*IV + … + parameter*IV
Y = β + β1X1+ β2X2 +… βnXn
There are many economic phenomena for which Linear in the Parameters (LIP) and Linear
in the Variables (LIV) models may not be appropriate or adequate (ex. Price elasticity of
demand, rate of growth of an economic variable such as GNP, money supply or
unemployment rate, etc.)

We will discuss 5 types of regression models that are capable of modelling such economic
phenomena:
1. Log-lin Model
2. Lin-log Model
3. Log-log Model
4. Reciprocal Model
5. Polynomial Regression Model

Linear regression analysis means that the parameters are linear that is, the maximum
power or exponential power of the parameters is one. Functional forms of regression
analysis are the models you adopt to represent the relationship between the independent
or explanatory variables and the dependent variable. It is important to choose the right
functional form for your regression analysis so that you avoid specifications in your model
and thus have a robust model.

Financial Econometrics | Functional Forms

 Page |3

I. Logarithmic Models
Considering the simple bivariate linear model Yi = α + βXi + εi, there are four possible
combinations of transformations involving logarithms: the linear case with no
transformations, the linear-log model, the log-linear model, and the log-log model.

I. No-log Model (Linear)

II. Linear-log Model
III. Log-linear Model
IV. Log-log Model

Note: Each of the model has its own advantages and disadvantages and practical
implications. We shall further study each model in detail. Since the linear model has
already been discussed, there’s no requirement to study it again.

Logarithmically transforming variables in a regression model is a very common way to

handle situations where a non-linear relationship exists between the independent and
dependent variables. Using the logarithm of one or more variables instead of the un-logged
form makes the effective relationship non-linear, while still preserving the linear model.

Logarithmic transformations are also a convenient means of transforming a highly skewed

variable into one that is more approximately normal. (In fact, there is a distribution called
the log-normal distribution defined as a distribution whose logarithm is normally
distributed – but whose untransformed scale is skewed.)

For instance, if we plot the histogram

of expenses, we see a significant right
skew in this data, meaning the mass
of cases are bunched at lower values.

Financial Econometrics | Functional Forms

 Page |4

On the other hand, if we plot the

histogram of the logarithm of
expenses, however, we see a
distribution that looks much more
like a normal distribution.

Thus, logarithmic aspect brings

down the scale of differences in
cases where the difference is just
too huge to measure and gives a
proper and significant conclusion.

❖ Linear-log Model: Yi = α + β(logXi) + εi

In the linear-log model, the literal interpretation of the estimated coefficient βˆ is that a
one-unit increase in logX will produce an expected increase in Y of βˆ units. To see what
this means in terms of changes in X, we can use the result that:

log X + 1 = log X + log e = log(eX)

In other words, adding 1 to log X means multiplying X itself by e ≈ 2.73. A proportional

change like this can be converted to a percentage change by subtracting 1 and multiplying
by 100. So, another way of stating “multiplying X by 2.73” is to say that X increases by
173% (since 100 × (2.73 − 1) = 173).

LEFT: Positive impact of Independent Variable | RIGHT: Negative impact of Independent Variable

After estimating a linear-log model, the coefficients can be used to determine the impact
of your independent variables (X) on your dependent variable (Y). The coefficients in a

Financial Econometrics | Functional Forms

 Page |5

linear-log model represent the estimated unit change in your dependent variable for a
percentage change in your independent variable.
Example:
An institute wishes to check if the fees paid by the students has any impact on the pass
percentage of the class or not. This is a time series data from 1960-2018.
Dependent Variable – PassPercentage
Independent Variable – Log(Fees)
The results calculated are below:

To interpret the coefficient of 4.851752 on the Log(Fees), we can make the following
statements:
• Directly from the coefficient: An increase of 1 in the Log(Fees) will increase Y by
4.851752
• A 1% increase in X: A 1% increase in Fees will increase Pass Percentage by
4.851752/100 = .0485
• A 10% increase in X: A 10% increase in Fees will increase Pass Percentage by
4.851752 ∗ log(1.10) = 4.851752 ∗ .09531 ≈ 0.4624
Thus, we can interpret that a further increase in the fees will have a positive impact on
the pass percentage of the institution. We can thus infer that higher fees makes the
student more diligent and self-aware of the cost and thus motivates himself to perform
better in the exams.

Financial Econometrics | Functional Forms

 Page |6

❖ Log-Linear Model: logYi = α + βXi + ui

In the log-linear model, the literal interpretation of the estimated coefficient βˆ is that a
one-unit increase in X will produce an expected increase in log Y of βˆ units. In terms of Y
itself, this means that the expected value of Y is multiplied by eβˆ. So, in terms of effects
of changes in X on Y (unlogged):
• Each 1-unit increase in X multiplies the expected value of Y by eβˆ.
• To compute the effects on Y of another change in X than an increase of one unit,
call this change c, we need to include c in the exponent. The effect of a c-unit
increase in X is to multiply the expected value of Y by ecβˆ. So, the effect for a 5-unit
increase in X would be e5βˆ.
• For small values of βˆ, approximately eβˆ ≈ 1+βˆ. We can use this for the following
approximation for a quick interpretation of the coefficients: 100 · βˆ is the
expected percentage change in Y for a unit increase in X. For instance, for βˆ = .06,
e.06 ≈ 1.06, so a 1-unit change in X corresponds to (approximately) an expected
increase in Y of 6%.

LEFT: Positive impact of Independent Variable | RIGHT: Negative impact of Independent Variable

After estimating a log-linear model, the coefficients can be used to determine the impact
of your independent variables (X) on your dependent variable (Y). The coefficients in a
log-linear model represent the estimated percent change in your dependent variable for
a unit change in your independent variable. The coefficient
Example:
If we reverse the previous case to check that whether the pass percentage of the class
leads to a change in the fee structure of the institution. The data and the time period is
same, from 1960-2018
Dependent Variable – Log(Fees)
Independent Variable – PassPercentage
The results calculated are below:

Financial Econometrics | Functional Forms

 Page |7

To interpret the coefficient of .1976864 on the urb95 variable, we can make the following
statements:
• Directly from the coefficient, transformed Y: Each one-unit increase of Pass
Percentage in increases Log(Fees) by .1976864.
• Directly from the coefficient, untransformed Y: Each one unit increase of Pass
Percentage increases the untransformed Fees by a multiple of e.1976864 = 1.054 – or
a 5.4% increase.

❖ Log-log model: logYi = α + β(logXi) + ui

In instances where both the dependent variable and independent variable(s) are log-
transformed variables, the interpretation is a combination of the linear-log and log-linear
cases above. In other words, the interpretation is given as an expected percentage
change in Y when X increases by some percentage. Such relationships, where both Y and
X are log-transformed, are commonly referred 0 to as “elastic” in econometrics, and the
coefficient of log X is referred to as an elasticity.
So, in terms of effects of changes in X on Y (both unlogged):
• Multiplying X by e will multiply expected value of Y by eβˆ

Financial Econometrics | Functional Forms

 Page |8

• To get the proportional change in Y associated with a p percent increase in X,

calculate a = log([100 + p]/100) and take eaβ

LEFT: Increasingly Positive impact of Independent Variable | CENTRE: Decreasingly Positive impact of
Independent Variable | RIGHT: Negative impact of Independent Variable

Example:
A survey was conducted to find out how much does the percentage of Urban Population
contribute to the Per Capita GDP of the nation. The survey was conducted in the United
Kingdom over a period of 59 years from 1901-1959.
Dependent Variable – LogGDP
Independent Variable – LogPop
The results calculated are below:

To interpret the coefficient of 12.51532 on the LogGDP variable, we can make the following
statements:
• Directly from the coefficient, transformed Y: Each one-unit increase LogPop in
increases GDP by 12.51532.
Financial Econometrics | Functional Forms
 Page |9

• Multiplicative changes in both X and Y: Multiplying X (GDP/cap) by e ≈ 2.72 multiplies

Y by e12.51532 = 0.9607, i.e. reduces the expected Urban Population by about 3.93%.
• A 1% increase in X: A 1% increase in GDP/cap multiplies Urban Population by e12.51532
∗ log(1.01) = .0.9950525. So, a 1% increase in GDP/cap increases Urban Population
by 0.5%.
• A 10% increase in X: A 10% increase in GDP/cap multiplies Urban Population by
e12.51532 ∗ log(1.1) ≈ .954. So, a 10% increase in GDP/cap increases Urban Population
by 4.6%.

❖ Reciprocal model: Yi = α + β2(1/Xi) + ui

When your dependent variable descends to a floor or ascends to a ceiling (i.e., approaches
an asymptote), you can try curve fitting using a reciprocal of an independent variable (1/X).
Use a reciprocal term when the effect of an independent variable decreases as its value
increases.
The value of this term decreases as the independent variable (X) increases
because it is in the denominator. In other words, as X increases, the effect of this
term decreases and the slope flattens. X cannot equal zero for this type of model
because you can’t divide by zero.

On the fitted line

plots, the quadratic
reciprocal model has
a higher R-squared
value (good) and a
lower S-value (good)
than the quadratic
model. It also
doesn’t display
biased fitted values.
This model provides
the best fit to the
data so far!

Financial Econometrics | Functional Forms

 P a g e | 10

Example:
Following is the study of the Infant Mortality Rate (IMR) and the Number of Hospital Beds
(per 1000). The data is cross-sectional taking 54 countries into account spanning over 5
continents.
Dependent Variable – IMR
Independent Variable – RHB (Reciprocal of Hospital Beds)
The results calculated are below:

As the number of hospital beds increases indefinitely, child mortality approaches its
asymptotic value of about 52 deaths per thousand. As explained, the positive value of the
coefficient of (l/RHB) implies that the rate of change of IMR with respect to number of
hospital beds is negative. Thus, this further solidifies this model as the increasing number
of hospital beds keeps having a negative impact on the child mortality rate.

Financial Econometrics | Functional Forms

 P a g e | 11

❖ Polynomial model: Yi = α + β2(1/Xi) + ui

The polynomial models can be used in those situations where the relationship between
study and explanatory variables is curvilinear. Sometimes a nonlinear relationship in a
small range of explanatory variable can also be modelled by polynomials.
The kth order polynomial model in one
variable is given by:
y = α + β1Xi + β2Xi2 + … + βkXik
The interpretation of parameter α is α =
E(y) when x = 0 and it can be included in
the model provided the range of data
includes x = 0. If x = 0 is not included, then
α has no interpretation. The polynomial
models can be used to approximate a
complex nonlinear relationship.

Example:
This is a study of a blue whale’s length and age. A dataset of 78 blue whales is recorded
and studied.
Dependent Variable – Length
Independent Variable – Age
Independent Variable – SqAge (Square of the Age)

Financial Econometrics | Functional Forms

 P a g e | 12

The results calculated are below:

This model is an example of a "second-order polynomial model" with one quantitative

predictor.
Length = 13.6 + 54.05Age - 4.719Age2
The R-Squared
value tells us that
80.1% of the
variation in the
length of blue
whale is reduced
by taking into
account a
quadratic
function of the
age of the fish.

CONCLUSION
There are a couple of advantages that non-linear regression has over linear regression,
which makes way for the functional forms. Each functional form comes with its perks and
practical applications. A clear disadvantage is that Linear Regression over simplifies many
real-world problems. More often than not, covariates and response variables don’t exhibit
Financial Econometrics | Functional Forms
 P a g e | 13

a linear relationship. Hence fitting a regression line using OLS will give us a line with a high
train RSS. Thus, functional forms help in applying differential weighting, identifying and
possibly removing outliers, and inspecting the correlation matrix and the dependency of
each parameter.
The table below summarises the whole assignment and briefs about the equation
structure, the slope and the elasticity.

SCOPE FOR FUTURE STUDY

The defining characteristic for both types of models are the functional forms. If you can
focus on the form that represents a linear model, it’s easy enough to remember that
anything else must be a nonlinear. Non-parametric techniques can fit any functional form,
but they typically take in all of the data when making predictions or they use some clever
method of reducing the data. There are further studies that can be conducted using
functional forms. Be it reciprocal and polynomial models in more than 1 variable, or a
further extension of the double-log model, the scope of future research is endless taking
into the account the extent of the robustness and the requirement of the academia.

SOURCES
1. https://statisticsbyjim.com/regression/difference-between-linear-nonlinear-
regression-models/
2. https://www.semanticscholar.org/paper/Analysis-of-Nonlinear-Regression-
Models%3A-A-Note-Peddada-
Haseman/a098591445d576f811f9e67057382dd9ccc8af34
Financial Econometrics | Functional Forms
 P a g e | 14

3. https://statisticsbyjim.com/regression/choose-linear-nonlinear-regression/
4. https://cmapskm.ihmc.us/rid=1052458916298_870839951_7777/Functional+form
5. https://www.graphpad.com/support/faq/advantages-of-using-prisms-nonlinear-
regression-analysis-to-fit-straight-lines/
6. https://books.google.co.in/books/about/Basic_Econometrics.html?id=byu7AAAAIA
AJ&redir_esc=y

Financial Econometrics | Functional Forms