Carrying out an Empirical Project

1
Carrying out an Empirical Project

• A researcher conducting an empirical study
  follows these basic steps
• formulate a model
• gather the data
• estimate the model
• subject the model to hypothesis testing
• interpret the results
• Model formulation-
• Single equation models
• Simultaneous equation models
• Examples-

2

• Hedonic price index model-
• in this model the price of a commodity depends on
  its characteristics. For example, we have a real
  estate agent who wants to relate the sales price
  of a house to its characteristics lot size,
  living area, number of bedrooms and bathrooms,
  types of built-in appliances, whether or not it
  has a swimming pool, whether it has a view, etc.
  In this model the marginal effect of a
  characteristic on the price is the shadow value
  of the characteristic.
• Simple hedonic model Price a bSqft cBaths
  dBedrms
• Capital Asset Pricing Model (CAPM)
• this model provides a general framework for
  analyzing risk-return relationships for assets.
  Let r be the return on a particular security, rm
  be the return on the market portfolio, and rf be
  the return of a risk-free return. Let Yr-rf be
  the excess returns of the market average
  portfolio. Then the following is the standard
  CAPM model
• Ybxu
• A security for which the beta is larger than one
  is more volatile than the market and one for
  which the beta is smaller than one is less.

3

• Data Gathering
• Model estimation
• Hypothesis Testing
• Interpretation of the Results
• Select a topic?
• One systematic way to approach the problem of
  choosing a specific topic is to utilize the
  classification system of the Journal of Economic
  Literature.
• Scan in table 14.1 (p. 652) from R
• Review the literature
• again the JEL is a good place to start
• Collection of the Data-
• where do you go?
• For data at the international level one can go to
  the International Financial Statistics (IMF),
  World Development Report (World Bank), Penn World
  Tables
• For data at the national level-US data from the
  BLS, NBER, Federal Reserve Board

4

• For data at the regional or state level there is
  state-by-state the Economic Report of the
  Governor which usually has statistical
  appendices. There is also the State and
  Metropolitan Data Book which is a supplement to
  the Statistical Abstract of the Unite States.
• Specialized data sources include the Survey of
  Current Business, often organized by SIC codes
  (standard industry classification). Federal
  Reserve Bulletin has detailed financial
  statistics. Standard and Poors and Moodys have
  ratings of securities and other financial
  information.
• Journals such as the Journal of Applied
  Econometrics, Journal of Business and Economic
  Statistics, American Economic Review, Journal of
  Money, Credit and Banking, have online archives
  of data used in empirical studies.
• Preliminary Empirical Analysis
• Box plots, check for the presence of outliers,
  summary statistics such as mean, median, standard
  deviations, coefficient of variation (one rule of
  thumb is to rule out variables whose CVlt.05), and
  well as the correlation matrix.
• If the data is a time series plot the series
  against time to get an understanding of possible
  periodicities as well as trend and growth rate
  properties.

5

• Model Estimation and Hypothesis Testing
• If you have enough data split the sample into two
  parts.
• The first will be used to analyze alternative
  specifications and examine significance of
  particular explanatory variables, functional
  form, etc.
• The second is to be used for your final
  estimations based on what you have learned from
  the first subsample.
• Next estimate the general model that you have
  formulated.
• Examine the fit (adjusted and unadjusted
  R-squared), significance and reasonableness of
  coefficients.
• Often this part of the analysis is complicated by
  poor initial results and may require a good bit
  of judgment.
• Diagnostics and formal specification checks for
  serial correlation, heteroskedasticity,
  nonlinearities, interactions, ommitted variables,
  and possible simulatenous equations bias can be
  assessed using a battery of Wald and Lagrange
  multiplier tests discussed in the Ramanathan
  text.
• Once you have a specification that has met with
  formal statistical criteria as well as with the
  smell test of reasonableness then reestimate
  using the second subsample
• Writing Up Your Results

6

• A suggested outline for writing an empirical
  report is as follows
• Statement of the problem
• review of the literature
• formulation of the general model
• data sources and descriptions
• model estimation and hypothesis testing
• interpretation of results and conclusions
• limitations of the study and possible extensions
• acknowledgements
• references and tables