Cecil fails to address what may have been the driving issue behind Tom’s question. For example, “Will it rain today?” is a particularly error prone area. Weather prognosticators currently lean on the percentages, but if they predict 20% and it rains on you all day, you’re thinking the prediction should have been 100%. I’m not defending the weather forecasters, just pointing out that “scattered showers” and “isolated thunderstorms” are real problem areas when it comes to forecasting accuracy. I don’t know if we can expect much improvement apart from sitting around watching the radar all day… 
Welcome to the Straight Dope duggles!
As it happens, the ForecastWatch study available here with a valid email address only reported temperatures, which are as the OP noted are less relevant than precipitation, alas.
Evaluating forecast accuracy along those lines might require some extra effort though.
A relevant question this raises:
Why do local stations run different weather forecasts when they get their data from the same expert, the government.
The local stations may have one meteorologist to the government’s five for their town.
And they certainly have no weather balloons or private Doppler (I don’t think).
Maybe they have a private rain gauge here or there, but why bother at all? They can only get farther from the experts.
You can see how accurate the weather forecasters are with their icons and text forecasts at ForecastAdvisor.com. Just type in your zip code, then click on “Further Analysis” on the bottom right. It will break down how accurate the forecasters are for high and low temperature and precipitation.
Also, for probability of precipitation, you might be interested in these two reports by ForecastWatch.com:
Accuracy of CustomWeather Probability of Precipitation Forecasts or (pdf)
Analysis of Short-Term Probability of Precipitation Forecasts or (pdf)
If you have any more questions, feel free to contact me. I’m the founder of Intellovations, LLC, which created both ForecastWatch and ForecastAdvisor.
Regards,
Ace
“Data” is different from “forecast”. Most meteorologists do get model information from the National Weather Service. However, there are many models, and they don’t really provide forecasts in the sense of something a viewer would understand. A good meteorologist uses the models, takes his own knowledge and experience, and mixes it all together to create a forecast. Some of the larger companies, like Accuweather, CustomWeather, The Weather Channel, or DTN/Meteorlogix will run their own computer models in addition to the government ones.
Actually, many midwestern television stations, especially in the tornado belt, buy their own Doppler radars. In fact, the “Doppler Wars” we get here have even been satirized on such shows as The Daily Show.
Regards,
Ace
Even Cleveland, a bit off the beaten track of Tornado Alley, has at least two local TV stations which boast of their Doppler radar arrays.
Plus, the national services don’t take into account critical data like the color of the wooly bear catapillers, or how long it took Phil to see his shadow. Or the status of the Weather Rock outside the meteorologist’s window. The truly good weathermen know how to account for these factors.