You are on page 1of 42

EC564 Financial Econometrics I (Time Series Analysis)

Stephen ONeill Department of Economics


St Anthonys Email: Stepheno_neill_1999@yahoo.com

What is Econometrics?

The study of methods that enables us to quantify economic relationships using actual data It includes all those statistical and mathematical techniques that are utilized in the analysis of economic data

A Stylisation of Econometrics Methodology (KP: 10-12)

Theory model
Observation of empirical regularities Assessment of past empirical and theoretical research Data instigated ideas/discovery

Outline formulation of a model to be estimated


Problems of matching theory and observable variables Empirical definitions of variables Collection of data and choice of sample period

Estimation of the empirical model


Choice of estimation method Estimation

A Stylisation of Econometrics Methodology (KP: 10-12)

Specification testing and diagnostic checking

Is the model statistically adequate? Does it pass diagnostic tests? Is respecification necessary? Is the model tentatively adequate?
Does the model improve upon previous models? Is the model able to encompass competing models?

Progressive evaluation of the empirical model


Asteriou and Hall Page 3

Economic Data

Economic data sets come in various forms which will influence the econometric methods applied Cross sectional Data [the focus for first few weeks!]

Consists of a sample of individuals, households, firms, countries, regions, cities or any other type of units at a specific point in time For each unit we have only one set of observation! Cross-sectional variables are denoted with the subscript i, taking values from 1,2,3,,N; for N number of cross sections, e.g. Y is national income for N number of individual countries. Yi for i = 1,2,3,..N

Example of cross section data (using made-up data)

Here we have 1 observation per country, for 3 variables GDP, population and some measure of natural resources. We can control for variation between countries but not for variation across time (since the data is for the same period)

Economic Data

Time Series Data [the main focus of this course!]

A time series data set consists of observations on one or more variables over time. We have multiple observations of the same unit Data are arranged in chronological order Data can take on different time frequencies annual/quarterly/weekly/hourly/ Examples include GDP, Stock prices, public expenditure etc. Time series data are denoted with the subscript t. So for example if Y denotes GDP of a country from 1995 to 2007 Yt for t = 1,2,3,,T. Where t = 1 for 1995 and t = T = 13 for 2002.

Example of Time series data

Here we have observations for only one country but we have an observation each year for 10 years. Here we can control for variation within a country but not for variation across countries (since we have only 1 countries data)

Economic Data

Panel Data

Panel data consists of a time series for each cross sectional member in the data set, e.g. Money supply data for a set of 20 countries over a 20 year periods. Panal data are denoted by the use of both the i and t subscripts. Hence money supply for a set of countries and for a specific time period is denoted by Yit for t = 1,2,3,,T and i = 1,2,3,..N

Example of Panel data

Here we have data for two countries for 5 years each. We can control for differences across countries but also for differences across time This is the most useful data to have, but not that common and estimation techniques may be more complex

Choosing between data types

Often the choice is made for you by unavailability of data! Otherwise, may be useful to think in terms of what you want from your results:

To predict for different individuals in the same time period => cross section To predict for a particular individual (e.g. country) in different time periods => time series To predict for different individuals in different time periods => panel data

For this course:

Although the focus of the course will be on time-series techniques, some knowledge of cross-section techniques/considerations is extremely important. We will spend some time on cross-section models. Unfortunately since this isnt the focus of the course it is your responsibility to get to grips with this part. Basic Econometrics by Gujarati gives a good (non-technical) intro to this stuff and I will give readings from this as we go. Please be sure to read them - particularly if you havent taken a course in econometrics before. Each year there is a clear difference between the marks of those who had done econometrics before and those that hadnt

For this course:

If you have any problems keeping up with the course e-mail me - I am happy to meet with any students that are struggling. If you do have problems:

Try to identify the specific area that is causing you difficulty:


Dont be general I dont understand ANY of it Be able to say I understood as far as . But where does come from?

There is a lot of terminology in econometrics but often the ideas arent as complicated as they appear. Try not to be daunted by it.

For this course:

After each class I will put up a list of questions similar to what tends to appear on the exam. I strongly recommend that people try to make sample answers to them

It saves time when it comes to the exam and makes you less likely to confuse topics More importantly it will help you to see whether you understand the topic.

I may make one question per lecture compulsory not as a hint that they will be on the exam - just so I can see if people understand what we cover

For the remainder of todays class:


We will look at some time series graphs Then well consider some definitions and commonly used data transformations

Time Series Graphs

A sequence of time series observations for a single (or several) variables recorded on a two dimensional graph with time on the horizontal and values of the observations on the vertical axis Why time series graphs are useful

Naturally order observations Provides a visual impression of the time series Observe tendencies to increase or decrease (trend) Observe tendencies to fluctuate more widely at different periods

Observe patterns

Time Series Graphs: Some Examples

From observing the time series graphs we see that economic observations are commonly dependent across time. Additionally time series data display clear trends over time. Time series data follow certain frequencies and might exhibit a seasonal pattern particularly in weekly, monthly and quarterly time series. We also observe structural breaks in series and also changes in the volatility of series. With structural breaks the time period of observation may matter. The frequency of observation also matters. Some series revert to a mean level while others demonstrate a random walk

Time Series Graphs: Some Examples

A time series typically consists of four components

Trend A smooth long term consistent upward or downward movement Cycle A rise and fall over periods, typically longer than a year Seasonal - within year patterns Irregular random component

Note: Not all time series have all components

See KP, Chapter 2 for further examples and discussion

Some Examples of Time Series Graphs [see excel sheets]

Irelands Nominal GDP 1970 - 1994


Ireland's GDP 1970-1994
50000
45000 40000

GDP in EUR, Current Prices

35000 30000 25000 20000


Irel

15000
10000 5000 De De De De De De De De De De De De De De De De De De De De De De De De De Year 0

We see a fairly smooth increase over the period If we look instead at the growth rate.

Growth rate of Irelands Nominal GDP


0.3

Ireland's Nominal GDP Growth Rate 1970-2006

0.25 0.2 0.15 % GDP Growth 0.1 0.05


Ireland

0 Dec-00 Dec-02
Dec-70 Dec-72 Dec-74 Dec-76 Dec-78 Dec-80 Dec-82 Dec-84 Dec-86 Dec-88 Dec-90 Dec-92 Dec-94 Dec-96 Dec-98 Dec-04 Dec-06

We see more variation. Nominal GDP includes the effects of prices Suppose we hold prices constant and look at Real GDP

Year

Irelands Real GDP 1960-2004


Irish Real GDP 1960-2004
120 100 80 Real GDP, 60 40 20 0
GDP

We see that there was a pretty smooth increase until 1994 and then it increased much more rapidly (but this was hidden by in the Nominal GDP series by lower price)

Year

12 10 8 % Growth 6 4 2 0 -2

Growth in Irish Real GDP 1960-2004

Year

We can see this also when we look at the growth of Real GDP growth is higher in the 1990s Also easier to see here that growth drops in the 2000s =>Important to know whats going on with your data -> is it telling you what you are interested in. If we are interested in productivity, its better to use Real GDP that to use Nominal GDP

Sometimes we observe periods where theres a lot of change (volatility), while at other times we see less change

ISEQ Daily level


ISEQ Daily Level May 2006- June 2007
10000 9500 9000 Index level 8500 8000 7500 7000

25.10

11.10

22.11

20.12

6500 10.5.06 24.5.06 21.6.06 19.7.06 16.8.06 30.8.06 13.9.06 27.9.06 7.6.06 5.7.06 2.8.06

28.3.07

8.11.06

6.12.06

17.1.07

31.1.07

14.2.07

28.2.07

14.3.07

11.4.07

25.4.07

Date

We see that at the start of the sample and towards the end, volatility is higher This is much easier to see when we look at growth rates.

23.5.07

3.1.07

9.5.07

Growth rate for Daily ISEQ


ISEQ Daily % Change 2006- June 2007
3.5 2.5 1.5 Index level 0.5 -0.5 -1.5 -2.5 -3.5 -4.5

So when you first work with a data series, its often a good idea to look at a few transformations of it to get an impression of whats going on!

Date

The importance of frequency of data.

Annual Tourism Revenue for Ireland


Annual Tourism Revenues in Ireland
1400 1200 1000 Tourism Revenue in 800 600 400 200 0

Dec-91

Dec-92

Dec-85

Dec-86

Dec-87

Dec-88

Dec-89

Dec-90

Dec-93

Dec-94

Dec-95

Dec-96

Dec-97

Dec-98

Dec-99

Dec-00

Dec-01

Dec-02

Dec-03

Dec-04

Dec-05

Year

We see a fairly smooth increase over time But if we instead obtain quarterly data

Dec-06

Quarterly Tourism Revenue


Quarterly Tourism Revenues in Ireland 1985-2007 2000 1800 1600 1400 Tourism Revenue in 1200 1000

800
600 400 200 0

Mar86

Mar03

Mar85

Mar87

Mar88

Mar89

Mar90

Mar91

Mar92

Mar93

Mar94

Mar95

Mar96

Mar97

Mar98

Mar99

Mar00

Mar01

Mar02

Mar04

Mar05

Mar06

Quarter

we see that although the overall trend is similar, there is much more variation, which may be important Whats causing this?

Mar07

Often our models work better when the data is less changeable, one way to achieve this is to take the log of the variable

Irish Real GDP 1960-2004


120 100 80 Real GDP, 60 40 20 0 Year
GDP

Natural Log Irish Real GDP 1960-2004


5 4.5 Change in Real GDP, 4
G

3.5 3

2.5

2
Year

Logs

Taking logs, just rescales the data However, we must remember that the results now relate to the log of the variable rather than the variable itself. => Important for interpretation

Summary Statistics and Definitions

A variable is something that can take on a number of values A discrete variable can take on a finite number of values [e.g. No. of rooms in a house] A continuous variable can take on an infinite number of values [e.g. Height] A random variable can take on a number of values each of which has a known probability [e.g. first number drawn in the lottery]

Summary Statistics and Definitions

Mean (average). This is a measure of central tendency 1 T

Mean:

x T
i 1

Variance: This is a measure of the extent to which a variable varies around its mean (A Measure of variability) T

Variance :

1 ( x x) T 1
2 x i 1 i

Standard Deviation:

2 x

Data Transformations

When using data to estimate an economic model a choice must be made as to how to apply the data. The frequency of the data series monthly/quarterly/annual

Annual data may obscure seasonal fluctuations Nominal series contain a price component which can obscure the fundamental features you are interested in

Nominal or Real values

Data Transformations

Logs

Linearizes an exponential trend which will reveal other features of the data series

Logs may be used to linearize a model which is non linear in the parameters. This is to enable the estimation of the model using standard techniques.

Consider the Cobb-Douglas production function.


Q AX Y log(Q) log( A) log( X ) log(Y )

Data Transformations

Differencing
Removes a trend component from a series entirely

Yt Yt Yt 1

(The change in Yt)

Although less common, we may difference a second time:

2Yt (Yt Yt 1 ) Yt Yt 1
(Yt Yt 1 ) (Yt 1 Yt 2 )

(The change in the change in Yt)

Growth Rates
Sometimes economists are more interested in the growth rate of a series rather than the level e.g Inflation Rate versus the Price Level Growth rate of Yt = (Yt-Yt-1)/ Yt-1

Lagging and Leading Time Series Data

A variable lagged s periods is denoted Yt-s


(I.e The value of Y, s periods ago.)

A variable leading k periods will be denoted Yt+k


(I.e The value of Y, k periods in the future.)

Time Series Econometrics

In time series econometrics the starting point is to obtain information from the variable itself Can analyse a single time series (univariate time series analysis) or multiple time series (multivariate time series analysis) In general the purpose of time series analysis is to capture and examine the dynamics of the data

Extract as much information as possible from the history of the data series

The Greek Alphabet:

From: http://www.cs.cmu.edu/~christos/christina/greece-project/alphabet.html

My lecture notes will be available on blackboard before class so you can print them out [set it to handouts rather than slides]

You might also like