Managerial Decision-Making Introduction To Using Excel In May 28-31, 2012 Thomas H. Payne, Ph.D. Dunagan Chair of Excellence in Banking Chair, Department of Accounting, Finance, Economics and Political Science
Preparing for : Installing the Analysis Tool Pak in Excel If you have Windows 7 and MS 2010 software on your PC: Open Excel and click on the File tab on the upper left hand side of the page, choose Options toward the bottom of the left hand side, then select Add-Ins. At the bottom of the page check Manage: Excel Add-Ins (in the drop down box) and click Go. From the dialogue box select Analysis Tool Pak & Analysis Tool- Pak VBA. That will do it!
The Importance of Demand Profitability of firms depends on the demand for the goods and services produced by the firm. In your case, it is important to have some reasonable prediction for the demand for deposits, loans, and fee generating services. Your home study project is designed to introduce you to the process of projecting that demand in the future. Note that this process should lead you to ask questions and better understand your data. As always, past trends are not always predictive of future results. However the better you understand those trends and the variables that affect them the better your forecasts will be.
Demand Time-Series models are useful for forecasting demand. Four core elements of a time-series model Long-Run Trends (LRT) Business Cycles (BC) Seasonal Variations (SV) Random Fluctuations (RF) Use data from the FDIC (www.fdic.gov) to forecast the number of farm loans in each quarter of 2011. Note that the example dataset is formatted and available on the GSB Website at http://www.gsblsu.org/3_8.html
Demand Long-Run Trends Linear Trend - 10,000.00 20,000.00 30,000.00 40,000.00 50,000.00 60,000.00 70,000.00 1986Q1 1986Q4 1987Q3 1988Q2 1989Q1 1989Q4 1990Q3 1991Q2 1992Q1 1992Q4 1993Q3 1994Q2 1995Q1 1995Q4 1996Q3 1997Q2 1998Q1 1998Q4 1999Q3 2000Q2 2001Q1 2001Q4 2002Q3 2003Q2 2004Q1 2004Q4 2005Q3 2006Q2 2007Q1 2007Q4 2008Q3 2009Q2 2010Q1 2010Q4 Farm Loans ($M)
Demand Long-Run Trends Let s start by controlling for a time trend Assuming a linear trend: y = a+bt Remember the equation that you used for a line back in school. The line intercepted the vertical axis at a. And the slope of the line, how much y changed for a change in x or in our case time (t), was b.
STEP 1 Add a new column on your spreadsheet and label it t. Then, for each line, enter the *year* of that observation. Time Period (As downloaded from the FDIC database) Loan Amount (your dependent variable i.e. what you eventually want to forecast) t (as provided in the dataset) 1986Q1 34,370.65 1986 1986Q2 34,690.13 1986 1986Q3 34,203.11 1986 1986Q4 31,602.33 1986 1987Q1 29,199.55 1987 1987Q2 30,819.61 1987 1987Q3 31,041.34 1987 1987Q4 29,427.25 1987
Demand Seasonal Variations But we can do better than this... What might we control for next if we are attempting to forecast farm loans in four quarters (time periods) of 2011?
1986Q1 1986Q4 1987Q3 1988Q2 1989Q1 1989Q4 1990Q3 1991Q2 1992Q1 1992Q4 1993Q3 1994Q2 1995Q1 1995Q4 1996Q3 1997Q2 1998Q1 1998Q4 1999Q3 2000Q2 2001Q1 2001Q4 2002Q3 2003Q2 2004Q1 2004Q4 2005Q3 2006Q2 2007Q1 2007Q4 2008Q3 2009Q2 2010Q1 2010Q4 Demand Seasonal/Quarterly Variation 70,000.00 60,000.00 Farm Loans ($M) 50,000.00 40,000.00 30,000.00 20,000.00 10,000.00 -
Demand Seasonal Variations But we can do better than this... What might we control for next if we are attempting to forecast farm loans in four quarters of 2011? Add a dummy variable for the season, quarter, or month of the year for which you have observations for your dependent variable: y=a+bt+cq1+dq2+eq3+fq4
STEP 2 Add four more columns representing the quarter of the year. Then, for each line, enter a 1 for the correct quarter, and a 0 otherwise. Time Period Loan Amount t Q1 Q2 Q3 Q4 1986Q1 34,370.65 1986 1 0 0 0 1986Q2 34,690.13 1986 0 1 0 0 1986Q3 34,203.11 1986 0 0 1 0 1986Q4 31,602.33 1986 0 0 0 1 1987Q1 29,199.55 1987 1 0 0 0 1987Q2 30,819.61 1987 0 1 0 0 1987Q3 31,041.34 1987 0 0 1 0 1987Q4 29,427.25 1987 0 0 0 1
Demand Observed Data Cycles One Final Consideration: Note that loans in 2004/2005 were off trend notice that the loans dropped off of the trend line during that time. We call this a *cycle*, and we want to control for it. For your bank project, the cycles you observe in the data may be lined up with true business cycles (i.e. recoveries and recessions). Here they are not however, another cycle does appear and is accounted for in the analysis. Note that our adjustment is observational but that observations typically have an underlying cause. So, you will want to consider things that cause this when analyzing your data.
1986Q1 1986Q4 1987Q3 1988Q2 1989Q1 1989Q4 1990Q3 1991Q2 1992Q1 1992Q4 1993Q3 1994Q2 1995Q1 1995Q4 1996Q3 1997Q2 1998Q1 1998Q4 1999Q3 2000Q2 2001Q1 2001Q4 2002Q3 2003Q2 2004Q1 2004Q4 2005Q3 2006Q2 2007Q1 2007Q4 2008Q3 2009Q2 2010Q1 2010Q4 Demand Business Cycle or other irregular data cycle. 70,000.00 60,000.00 Farm Loans ($M) 50,000.00 40,000.00 30,000.00 20,000.00 10,000.00 -
STEP 3 Add one more column representing the observed cycle. Again, code this as a 1 for years that are in the cycle, and a 0 for years that are out of it. The series below shows part of this cycle which we estimated as lasting from the 2 nd quarter of 2003 through the 4 th quarter of 2005. Time Period Loan Amount t Q1 Q2 Q3 Q4 Cycle_1 2005Q1 45,379.52 2005 1 0 0 0 1 2005Q2 48,273.03 2005 0 1 0 0 1 2005Q3 50,707.90 2005 0 0 1 0 1 2005Q4 51,669.39 2005 0 0 0 1 1 2006Q1 49,242.74 2006 1 0 0 0 0 2006Q2 52,706.24 2006 0 1 0 0 0 2006Q3 54,009.98 2006 0 0 1 0 0 2006Q4 54,256.92 2006 0 0 0 1 0
Demand Now the full equation for our farm loan model looks like: y=a+bt+cq1+dq2+eq3+fq4+ gcycle_1+ RF
STEP 4 Run the full model in Excel to calculate the values of a, b, c, d, e, f, & g. Steps: Data -> Data Analysis -> Regression -> OK Select Y-Range: B1-B101 Select X-Range: C1-H101 Check the Labels box Pick an Output Range (New Sheet) Coefficients Intercept -2470997.225 t 1257.517677 Q1 0 Q2 2787.00352 Q3 3670.70032 Q4 3059.24012 Cycle_1-3987.266825 Note: One of your four quarters (or 12 months) will be zero. In our examples it was quarter 4 (the one we did in class) or quarter 1 (the example above); regardless, however, the answer will be the same when you plug the appropriate data and associated coefficients into your forecast model.
Demand Predicting Future Values of y Now we have a formula that tells us the relationship between y (farm loans) and my control variables. Y = -2470997.2 + 1257.52 x t + 0 x Q1 + 2787.00 x Q2 + 3670.70 x Q3 + 3059.24 x Q4 + -3987.27 x CYCLE1 Coefficients Intercept -2470997.225 t 1257.517677 Q1 0 Q2 2787.00352 Q3 3670.70032 Q4 3059.24012 Cycle_1-3987.266825
Checking Our Results Your home study project will allow you to predict loans, services, etc. through late 2012 and into 2013. However, for example purposes, this example used data through 2010 to predict farm loans into 2011. The results are shown on the panel on the right. Farm Loans Predicted in 2011 Actual Farm Loans in 2011 (Millions) Year/Quarter Predicted (In Millions) $61,337 2011 (Q4) $56,947 $59,802 2011 (Q3) $61,546 $57,668 2011 (Q2) $60,663 $55,033 2011 (Q1) $57,876 Error (predicted-actual) Percent Error -$4,389-7.16% $1,744 2.92% $2,995 5.19% $2,842 5.16% Y (Loans) = -2470997.2 + 1257.52 x t + 0 x Q1 + 2787 x Q2 + 3670.70 x Q3 + 3059.24 x Q4 + -3987.27 x CYCLE1