Dear Manisha,

When I was working for the US Dept. of HUD, we were faced with a similar problem in forecasting. The methodology prescribed by the "FHA Handbook" is to make an estimate for the current period. In your case you have census data up to 2011, and trend for 1991 to 2001. Use the census trends, along with local estimates for the year 2014, and whatever stable ratios you can find such as participaiton rates, household size, institutional population, capital/output ratio, etc. to estimate the 2014 value. 

I mention only HUD's technique because it was officially used by a government agency. There are other judjemental techniques you may wish to explore.

Gook luck in your research.

ya its official census data. Its conducted every 10 year so I have it just for three years i.e. 1991, 2001, 2011.

Can you suggest me some good books or articles the can help.

Thanks a lot.

A Simple and readily usable method is that one can use the actual growth rate between the two Census years i.e. 2001 and 2011 to interpolate the numbers for any year between 2001 and 2011 and then extrapolate to any year after 2011 but not for too long years, assuming that rate of growth observed during the period 2001 and 2011 will be continued. This formula may work.  Pt = P2001 * (1+r)n ; where Pt is any year for which we want to interpolate or extrapolate, P2001 is the base year, r - rate of growth; n - number of years between the base year and the year for which we want interpolated/extrapolated number. 

i need data for project, it is collected by the government through a survey that is conducted every 10 years. so i have data for 1991, 2001 and 2011. But my study requires yearly data, so how should I proceed?

Interpolation/extrapolation using related series (that are available at higher frequencies) might be useful here. Between decennial surveys, the US Census Bureau publishes annual population estimates which are then used at a fairly high frequency (monthly) to estimate levels of labor force and employment. I don't know how much related data or alternative high frequency series you have available for this work; it will be important to use as much higher-frequency data/information as possible I think.  Good luck!

...one of the better known and influential works in this area was published in 1971 by Chow and Li: Best linear unbiased interpolation, distribution, and extrapolation by related time series. There have been subsequent works, but the original Chow-Li is a good place to start. Procedures to do this kind of thing are fairly common. However, the paucity of annual observations constrains the range of possible inference here.

http://www.oecd.org/std/21779751.pdf

Can you also help me with some good papers where no related series are available

Hi Manisha. When there are no related series (an unusual situation, I think) there are limits on what can be done. Simple linear/polynomial interpolations or extrapolations might be used to fill in the missing years, but we'd have no way of knowing whether arbirtrary choices between ad hoc methods are truly adding information. It's hard to imagine there are no data available at all for the missing periods that might be used.