When we record employment levels over a time period, we observe following five distinct elements in it.

Trend, which is the fluctuation in level of employment over a time period.

Cyclical effects that are the changes in employment in relation to some particular event, like economic liberalization in India or WTO resolutions.

Seasonality, which is a seasonal fluctuation that occurs more than once in a given time period, like requirements of more maintenance staff in Delhi Vidyut Board during summer and rainy season while less in winter.

Step is a sudden change in employment level due to economic environment or increased market share or procurement of some new machines, etc.

Random fluctuations are the fluctuations in employment level which are random in nature, i.e., such fluctuations do not follow any obvious pattern. This problem can be addressed by following a Moving Average Method.

Some illustrations of manpower forecasts using Time Series Analysis are given in the following.

#### 1. Moving Average Method:

Under this method, the average of the combined employment level data for the recent past is considered as the forecasted employment level for the next period. This technique is useful to guard against random fluctuations. However, it requires careful selection of time periods, which may be a 6-period, 10-period and 12-period moving, average.

If time periods chosen are too few, we may get wide variability in our forecast. Therefore, considering more time periods, we get better results. Whatever time periods are selected for forecasting, it is necessary to continue the same number of periods for our computations. After each period elapses, the figure for the oldest period is dropped and the figure for the newest period is added for the subsequent computation of manpower requirements.

Example:

Godrej has the following manpower data for their health care division for the past six years:

Year

Manpower Level/Data (in nos.)

2004

500

2005

600

2006

800

2007

1000

2008

1100

2009

1300

You have been asked to forecast their manpower requirement in 2010 using a6-period moving average.

Solution Fm = 500+600+800+1000+1100+1300/6 =5300/6=883nos.

If we are asked to use a 4-period moving average to forecast manpower for 2010, then we are required to drop the data of 2004 and 2005 and compute the forecasted manpower as under:

Fm = 800+1000+1100+1300/4 =4200/4 =1050 nos.

From the above variation in results, we can well understand the danger of considering lesser time periods in manpower forecasting.

For achieving better results, weights may be assigned for different time periods at the discretion of the analyst. To take an example, the manpower level of 2004, 2005 and 2006 may be less relevant, hence, for these years weights may assigned as 1 each. For 2007, the weight may be 2, for 2008 the weight may be 3, whereas for 2009, the weight may be 4. Now, the forecasted manpower for the year 2010, would be as under:

#### 2. Exponential Smoothing:

In Moving Average Method, it is need to carry forward large volume of historical data. The need for such past records can be eliminated adopting this method. This method smoothens random errors by giving exponentially decreasing weights to historical data.

Such weight factor is indicated by alpha (?), which is a smoothing constant, a non-linear decimal value which lies between 0 and 1. The formula for the exponential smoothing model is:

Fm = Ft + ? – (At–1 – Ft- l)

Where,

Fm = Forecasted manpower

Ft 1 = Forecasted demand for the previous period

? = Smoothing constant

At-1 = Actual manpower required for the previous period

Example:

Assume that the forecasted manpower requirement for an organization was 500, whereas, their actual requirement was 480. Considering alpha value of 0.4 (which the company feels would produce the best results), compute the manpower requirement for the current period.

Solution: 500 + 0.4 (480 – 500) = 492 nos.

Double exponential smoothing can also be done to get further cushion in manpower forecasting.