ECON5248 Business Forecasting

The generation of new business start-up is vital to the growth of the economy as it builds new jobs and creates new opportunities for the community. The Bureau of Labor Statistics tracks new business development and jobs created on the website for the United States Department of Labor. You have been tasked with forecasting economic growth and decline patterns for new businesses in the United States. 

Answer:

Summary

This paper sought to conduct a forecasting of the number of establishments that are less than 1 year old for each year during this period. We selected the 5 most recent years (2001 to 2005). Four different forecasting methods were employed, they include;

• Exponential Smoothing

1. Trend Adjusted Exponential Smoothing

Each forecasting method was analyzed on its work sheet.

Several statistics were employed used to compare the fits of the different smoothing and forecasting methods. The statistics used include;

1. Mean absolute percentage error (MAPE)
2. Median absolute deviation (MAD)

• Mean square error (MSE)

1. Bias

Mean absolute percentage error (MAPE)

This statistic measure the accuracy of the forecast as a percentage of the error.  The method is much easier to comprehend as compared to other methods (Makridakis, 1993). For instance, if MAPE is given on average to be 3, then this tells that the forecast is off by 3%. The smaller the percent the MAPE has, the better the forecast. The MAPE computation formula is;

Where = actual value at time t, =fitted/forecast value and n = number of observations.

Median absolute deviation (MAD)

MAD is a robust measure that identifies how spread out the given data is. MAD is different from other measures of spread in the sense it is not that much affected by extremely low or extremely high values and non-normality (Hyndman & Anne, 2006). MAD is best to measure non-normal datasets.

There is also the mean absolute deviation (MAD) that also helps conceptualize the amount of error. It is also not that much affected by the outliers.

Where = actual value at time t, =fitted/forecast value and n = number of observations.

Mean squared error (MSE)

This is a commonly used measure of accuracy for the fitted time series. It is also known as mean squared deviation (MSD). This statistics is however greatly affected by the outliers (Tofallis , 2015). The formula for computing MSD is given as follows;

Where = actual value at time t, =fitted/forecast value and n = number of observations.

Trend analysis

As can be seen in the figure below, the number of establishments that are less than 1 year old have been on a rise. It is only in 2013 when the numbers slightly went down.

The rise in the number of establishments means that the number of emerging and new entrepreneurs has been on rise.

Forecasting methods

As mentioned earlier a number of forecasting methods were employed to forecast the number in the coming next two years. Different measures of accuracy were also computed to identify the best forecasting method. The table below gives the results of the same;

Table 1: Measures of accuracy

Review of error

From the table presented above (table 1), it is evident that trend adjusted exponential smoothing has the lowest measures of accuracy. The lower the accuracy the better the forecast method. The MAPE for the trend adjusted exponential smoothing is 3.65% while that of the 3-period moving average is 5.97%. This implies that forecast are off by 3.65% and 5.97% respectively for the two forecast methods. Thus, the best forecast method out of the four methods would be the trend adjusted exponential smoothing followed by the 2-period moving average while the 3-period moving average would come last.

Works Cited

Hyndman, R. J., & Anne, B. K. (2006). Another look at measures of forecast accuracy. International journal of forecasting, 22(4), 679-688.

Makridakis, S. (1993). Accuracy measures: theoretical and practical concerns. International Journal of Forecasting , 9(4), 527-529.

Tofallis . (2015). A Better Measure of Relative Prediction Accuracy for Model Selection and Model Estimation. Journal of the Operational Research Society, 66(8), 1352-1362.