Complexity is one answer: small errors accumulate and become amplified as they work through a forecasting model. Data is another: third quarter GDP growth numbers won't be available until the end of November, and they will be revised five times before the final numbers are published in May, 2014 (revisions of half of a percentage point are typical). These and other issues pose significant technical problems, and progress is slow in dealing with them.
Forecasters who work outside academia don't have the time or resources to give more than cursory attention to these issues, so they rely on a mix of models and subjective opinion to produce their projections. There's nothing wrong with using intuition in a forecast, but its contribution should be transparent.
Using the average of private-sector forecasts makes sense if private-sector forecasts are being produced independently. But instead of trying to produce forecasts that match the data, private forecasters seem to be trying to match each other. One study notes that “[t]he range of forecasts underestimates the degree of uncertainty facing forecasters, sometimes substantially.”