basic 3 year's observations forecasting - need for tips and feedback



Hello everybody,

I need advice about simple forecasting strategies and wondered if anyone here could kindly give me feedback. In each of the four charts shown below we have three results (2011 to 2013) of the number of farmers (above) and school teachers (below) that have enrolled and received training as part of a large-scale basic ICT use program.

It is now required to have a forecast of results for the two subsequent years for planning purposes. To this end, I used Excel to draw a regression trend line (linear on lefts) and polynomial on rights where we can visually see the how things might turn out.

The problem is that using the forecasts on the right gives unrealistic growth prospects in spite of being the best fit. There is no way 1 million farmers will enroll by the fifth year and what scares me most, -34 000 teachers enrolled and taking courses by the fifth year is totally out of order.

If I use the linear regression version, the fitness of the trendine R-squared is much lower than if I took polynomials but the predicted values for year four and five seem much more real and meaningful.

What would you do? do I just have too little data to make a forecast? Or should I rather base my forecast on another method as the mean of previous observations?

Many thanks for any input on this.