ECON 300 Forecasting Projects

How to format a microtheme.

Grading rubric.

Assignment #1 CONSUMER PRICE INDEX FORECAST

Part 1: Technical Component

Estimate two equations for the CPI using lagged M2 as the explanatory (right-hand-side, RHS) variable.  For your first estimate and forecast, lag M2 one year, and for the second lag it two years.  Use data ending at December 2008 in your regressions and use the estimated equations to compute a forecast of CPI for the twelve months of 2009.  Compare your forecasts with the actual CPI for the available months of 2009 by computing the root mean square error (RMSE)—the forecast with the lowest RMSE is the best.  In addition to your paper presenting and interpreting the forecasts, hand in spreadsheets with the original data on one worksheet, each regression model, its accompanying forecast, and the RMSE computation together on its own worksheet.  (Your Excel workbook should have three sheets.)  These can be emailed to me or handed in in-class as printouts.  Data can be found at the FRED website: http://research.stlouisfed.org/fred2/

Part 2:  Presenting and communicating the Forecasts

The microtheme will be due the Monday of the week prior to the due date of the finished paper.  Prepare a three-page report presenting and discussing your forecasts.  The paper will include a chart illustrating the two forecasts, and two tables, one listing the RMSEs for the forecasts, and one listing the actual forecast values (from your second, reestimated, larger-sample regressions).  Note that the second forecast will be able to extend twice as far into the future as the first.  This report and the microtheme will be graded according to the rubric given above.

Assignment #2 FORECASTING UNEMPLOYMENT AND GDP WITH THE PHILLIPS CURVE

Part 1: Technical Component

Download data for the CPI and unemployment from FRED.  The url is: http://research.stlouisfed.org/fred2/.    Download these data files: CPIAUCSL and UNRATE.  UNRATE is under the Household Survey link of the Employment and Population link.  Use CPIAUCSL to calculate the inflation rate as INF(t) = [CPIAUCSL(t) – CPIAUCSL(t-12)]/CPIAUCSL(t-12).  Lag INF and UNRATE 12 and 24 months each.  Estimate three regression equations for UNRATE: (1) with INF(t-12) and UNRATE(t-12) as the explanatory variables, (2) with INF(t-24) and UNRATE(t-24), and with INF(t-12), INF(t-24), UNRATE(t-12), and UNRATE(t-24).  Make sure the sample period for these regressions ends in December of 2008.  Construct the forecasts as far into the future as possible given the available data, and calculate the RMSE for each.  The RMSEs will be reported in your paper to evaluate the forecasts.

Then go back and reestimate the equations using the all the more recent available observations for 2009 and 2010, and construct new forecasts as far into the future as possible.  Report each forecast in a table for the relevant period.

The number of months of data should be the N reported in the regression output for all three regressions.

Part 2:  Presenting and communicating the Forecasts

The microtheme will be due the Monday of the week prior to the due date of the finished paper.  Prepare a three-page report presenting and interpreting your forecasts.  The paper will include a chart illustrating the three forecasts and two tables, one listing the RMSEs for the forecasts, and one listing the actual forecast values (from your second, reestimated, larger-sample regressions).  Note that the second forecast will be able to extend twice as far into the future as the other two.  This report and the microtheme will be graded according to the rubric given above.  Your thesis statement should reference the range between your largest and smallest forecast values for inflation.

Assignment #3 FORECASTING GDP

Part 1: Technical Component

Download data for Real GDP, its Chain-type Price Index, and Potential Real GDP.  The url is:
http://research.stlouisfed.org/fred2/.   Download these data files: GDPC96, GDPCTPI, and GDPPOT.  On GDPCTPI, calculate the average for the four quarters of 2000, divide all values by this average, and multiply by 100.  This converts the base year from 2005 to 2000.  Multiply each value of GDPPOT by the base 2000 index, and divide by the original base 2005 index.  This converts Potential Real GDP measured in 2000 dollars to 2005 dollars, making it comparable to Real GDP.  Estimate an autoregression of Real GDP on a lagged value of itself, and use the estimated coefficients to project a forecast to the fourth quarter of 2019.  In addition to the regression output normally provided by Excel, compute the AIC, SBIC, and the LogLikelihood.

You will each be given a different order of autoregressive process to implement:
estimate an AR(2) regression: Yt = a + b(Yt-2) + et
estimate an AR(3) regression: Yt = a + b(Yt-3) + et
estimate an AR(4) regression: Yt = a + b(Yt-4) + et
estimate an AR(5) regression: Yt = a + b(Yt-5) + et
estimate an AR(6) regression: Yt = a + b(Yt-6) + et
estimate an AR(7) regression: Yt = a + b(Yt-7) + et
estimate an AR(8) regression: Yt = a + b(Yt-8) + et.

Use the coefficient standard errors to construct upper and lower one-standard-error forecast limits.  Graph your forecast values, the standard errors, and potential GDP.

For each quarter, calculate the percent difference between potential and forecast Real GDP, and use Okun's Law to estimate unemployment.  Okun's law states that unemployment will be 1% above the natural rate for every 2% actual GDP falls below potential GDP.  Assume a natural rate of unemployment of 6%.  How does this change if you assume a natural rate of 3% or 4%?

Part 2:  Presenting and communicating the Forecasts

The microtheme will be due the Monday of the week prior to the due date of the finished paper.  Prepare a three-page report presenting and interpreting your forecasts of Real GDP and unemployment.  The paper will include a graph illustrating the forecast with its upper and lower standard errors, and potential Real GDP. This report and the microtheme will be graded according to the rubric given above.  Your thesis statement should reference the highest value forecast for Real GDP, and whether this is above or below Potential Real GDP for the corresponding quarter.  A topic sentence and paragraph can refer to the unemployment forecasts. 

Assignment #4  FORECASTING RECESSION WITH THE FINANCIAL INSTABILITY HYPOTHESIS

Part 1: Technical Component

The attached spreadsheet contains data from Compustat for around 8000 publicly traded equities.  Using their NAICS codes, segregate out the industry groups you are assigned below, and delete all the other equities.  (Cut and paste the whole block of data onto two different worksheets, and keep the original worksheet if you want to keep the original data for your own uses.)  When you delete the data for all other industrial groups, the formulas at the bottom of the page will automatically adjust to recompute Minsky's Financial Instability Hypothesis categories for each quarter of data.  Make sure you understand the formulas in each cell.  The graphs will have to be redone, but you can use the existing graphs as templates, removing the original variables and replacing them with the corresponding ones you compute for your industry groups.

The North American Industrial Classification System (NAICS) assigns each company a 2-6 digit code beginning with the following two digits for each industry.

Each student will be assigned the industry groups to use:

Transportation & Warehousing 48-49

Information 51
Professional, Scientific, & Technical Services 54

Management of Companies & Enterprises 55
Agriculture, Forestry, Fishing, & Hunting 11

Mining, Quarrying, & Oil & Gas Extraction 21
Construction 23

Real Estate, Rental, & Leasing 53
Education
61

Health Care & Social Assistance 62
Retail Trade 44-45

Accommodation & Food Services 72
Utilities 22

Manufacturing 31-33

For each NAICS industry group, compute the percent of listed firms which are either Speculative or Ponzi Finance Units.  Regress this number on two-year-lagged quarterly interest rates (3 month Treasury bill, secondary market interest rate).  You’ll find an example of this regression with the quarterly interest rate series on the “Minskyinterest” spreadsheet.  Use this regression to construct a forecast two years into the future.  Use the coefficient standard errors to construct upper and lower one-standard-error forecast limits. Graph your forecast values and the upper and lower standard error bounds.  In addition to the regression output normally provided by Excel, compute the AIC, SBIC, and the LogLikelihood.

Part 2: Presenting and communicating the Forecasts

The microtheme will be due the Monday of the week prior to the due date of the finished paper.  Prepare a three-page report presenting and interpreting your forecasts of the number or percent of firms in each sector which will be non-Hedge-finance-units, and state whether this is consistent with recovery or continued recession.  The paper will include two graphs illustrating the percent Speculative and Ponzi-finance-units in each NAIC industry you were assigned to cover, with the forecast and its upper and lower standard errors.  This report and the microtheme will be graded according to the rubric given above.

Include any other graphs you think are interesting. 

Assignment #5  FORECASTING SUPPLY AND DEMAND WITH INSTRUMENTAL VARIABLES

Part 1: Technical Component 

The attached spreadsheet contains data pertaining to the market for copper.  Use two-stage least-squares to estimate unbiased supply and demand functions.  Assume the following functional forms:

DEMAND: QCOPPER = a + b(PCOPPER) + c(INCOME) + d(PALUMINUM)

SUPPLY: QCOPPER = f + g(PCOPPER) + h(INVENTORY) + i(TECHNOLOGY)

First estimate the first-stage equation PCOPPER = j + k(INCOME) + l(PALUMINUM) + m(INVENTORY) + n(TECHNOLOGY)

Ask for the residuals.  The fitted value of this equation will be given next to the residuals and labled "Predicted PCOPPER."  Use this in place of the real PCOPPER in your Demand and Supply regressions.  These will be the second-stage equations.

Use your estimated coefficients from the supply and demand regressions to compute tables of elasticities.  Interpret them in your paper.

The attached spreadsheet has the data on sheet one and an example like the one done in class on sheet two.  Use this as a template, but make sure you do your own regressions, because I changed some of the data.  Notice that all I did was change a few of the income numbers, and it gives completely haywire signs—both the supply and demand curves have the wrong signs!

Part 2: Presenting and communicating the Forecasts/Elasticities

The microtheme will be due the Monday of the week prior to the due date of the finished paper.  Prepare a three-page report presenting and discussing your estimates of the elasticities.  You can emphasize either the most recent elasticities or the averages.  This report and the microtheme will be graded according to the rubric given above.

The final report and the Excel spreadsheet are due in the final exam period.  (I’ll be in my office if I’m not in the classroom.)