FORTUNE -- Making a forecast for one month's car sales is like trying to predict the weather: It's extremely tricky, and nobody pays attention to it unless you are wrong.
Looking at longer-term auto sales trends is more like projecting changes in the climate: All the ingredients are visible, and the findings resonate with statistical clarity.
Never has the dichotomy been clearer than now.
When two respected forecasters got out their crystal balls to gauge the direction of September auto sales, they came up with sharply different conclusions -- even though the end of the month is only days away.
Meanwhile, a five-year forecast by a third organization came up with another set of numbers that seemed far more robust -- and, just incidentally, should gratify everyone in the industry.
First out was J.D. Power & Associates, the industry's longtime product and market analyst. Seven days before the end of the month, it declared that September vehicle sales were expected to post "a significant selling rate increase" compared with August.
J.D. Power, which gathers real-time transaction data from more than 8,900 dealers, figured that the annualized selling rate for September would be 11.8 million units, a healthy increase over August's rate of 11.4 units.
Opined veteran forecaster Jeff Schuster: "The strength in the first half of September is exactly what the industry has been looking for to begin a more measurable recovery through the reminder of the year with continued progress into 2011."
Just a few hours later, Edmunds.com, the up-and-coming consumer website, was out with a different forecast. Edmunds, a broad-based source of product information for shoppers has been trying to establish itself as an alternative source of auto data as well by providing instant analysis of industry events.
Edmunds figured that September sales would actually fall 5.5% from August, and the annual selling rate would be essentially flat: 11.47 million vehicles in September vs. 11.44 in August.
While Power saw the glass as half-full, it looked half-empty to the analysts at Edmunds.
"Despite some noteworthy new car introductions, auto sales are stagnant right now," Edmunds' senior analyst Jessica Caldwell said in a statement. "Automakers seem to be reluctant to invigorate the market through traditional incentives programs or unload significant levels of inventory as fleet sales."
It won't take long to determine who's right. September sales results will be announced October 1.
Both forecasts predict rather unimpressive short-term results. But another forecast that peers farther out has much rosier view.
Looking at the long-term picture is easier because U.S. car sales are driven by two easily-measurable factors: the growth in the number of licensed drivers and the scrappage rate of older vehicles. Cars and trucks are just consumables with a longer life span. Replacement can be deferred for a time but not indefinitely.
In a report dated August 30, analysts from Morningstar Institutional Equity Research make the case for sales quickly reaching an annual sales rate of 20 million vehicles -- nearly double the current rate.
After a recession such as the one the economy has undergone, Morningstar says the effect on auto sales is like a coiled spring. When sales fall below the usual trend for a long period of time, as they have for the last two years, they have to spring above normal rates so that the worn-out vehicles that have remained on the road can be replaced.
After running the numbers, such as the percentage of existing drivers buying a new vehicle, Morningstar detected a boom in auto sales just over the horizon. Rather than poking along at 11 million vehicles a year, it believes that demographics and wear-and-tear will require the sale of 50% more cars and trucks almost immediately.
It is figuring that pent-up demand that has been building since 2007 will require a sales rate in excess of 17.4 million vehicles in 2011 and an annual rate of far more than that.
"We believe that over the next five years, the growth in the pool of licensed drivers and cumulative pent-up demand will drive U.S. light-vehicle sales to as total 96 million units, or an average run rate of more than 19 million units," says the report.
The conclusions reached by Morningstar will shock the majority of forecasters who say car sales are dependent on consumer confidence, and consumer confidence won't recover until the unemployment rate, now north of 9%, starts to decline.
It will also jar theorists who believe that booming car sales earlier in the decade (when annual sales did reach 17 million) were artificially stimulated by rising house prices and marketing incentives driven by over-production, and won't be repeated.
If the Morningstar forecast is accurate, it will mean a bonanza for domestic automakers. GM will reap a fortune with its initial public offering, Ford (F, Fortune 500) can restore the dividend on its common stock, and Chrysler can solidify its plans for the future.
But it will also put a huge strain on resources. Automakers downsized their production capacity through the recession and would be hard-pressed to meet soaring demand. That would leave an opening for foreign automakers, including those from China and India.
Is the Morningstar analysis on the mark? The authors admit their model "over predicted" the size of the recovery from the 1981-82 recession.
But they add, "The main conclusion that we can draw from this model is that the effect of a return to normalized demand, plus an increase in sales to fill pent-up demand should result in a large increase in new light-vehicle sales within the next few years."