impact significance of each impacts of intervention will calculate the exist measurement of impacts.

Quote from 

http://www.ifad.org/evaluation/public_html/eksyst/doc/impact/2013/srilanka/dzlspp_impact.pdf

One of the fundamental constraints in the context of this

evaluation was the absence of a baseline dataset. For this reason, the quantitative

component of the survey employed two strategies: (i) an attempt to reconstruct

baseline information through recall methods; and (ii) a quasi-experimental approach

using statistical techniques that do not strictly require baseline data. In particular,

the evaluation adopted “propensity score matching” as well as the “treatment effect

model” (an application of the Heckman sample selection model) to test for

consistency and robustness of results. Both can help address sampling bias when

project participants have not been randomly selected.

Dearest MD, where baseline data are absent, the best ways to measure the impact of intervention is to select short term indicators such as changes in adoption rate of the intervention, yield of adopters, income of adopters compared to laggards, and knowledge gained after the intervention, in fact you can also ascertained if intervention has affected the skills and knowledge of adopters and how this will impact the community and food security.

Where basline data are not available, your own data will set the base line for comparisons after your study. Employing a statistical technique did not reuire a baseline data. For example, you are emplying an experimental treatment for the first time. The control (Without this treatment) will be your baseline data for your comparison. Then your data of this treatment will prove a baseline data for others under similar set of conditions.

You may use the quasi-experimental approach - with (treatment group) and without (as your baseline. You may also consider the Retrospective Pretest Model where you  administer the preprogram assessment concurrently with the posttest by asking individuals to recall their knowledge or behavior prior to the program.

you need to design an impact assessment study/survey that tries to estimate the changes the beneficiaries have experience over the time of intervention

Since baseline data are not available you cannot estimate directly the impact of the change variable on the sample. What you might do is divide the sample, if possible, into two groups: one that was subjected to the change variable and the other that did not and use the difference as an estimate of the impact of the change variable. However, this will not capture the effects of individual specific factors if any.

In the absence of baseline, the IA can be conducted using "with and without" approach. However, you need to be careful in selecting the group without intervention. The approach can be combined with quasi "before and after" approach by asking the selected respondents to recall the condition before the intervention. For example, if the project intervention is in, said, 2012, then both group can be asked to recall the condition in 2011 and 2013. The combination then will be similar to "different in different" approach. Any differences in indicators you selected as the impacts between the groups can be attributable to the project intervention. Hope this method is useful. 

Comparison of with and without intervention is sometimes used as a method but that is not fully appropriate. Recall baseline data can be collected, as some respondents have suggested. This is is not perfect but better than not having any baseline. After all the purpose is to estimate the general pattern of change or impact, not the unique values of any change as no sample  based estimate is fully accurate.

For a general guideline, see relevant section in the attached paper.

Hi, You can adopt Experimental Research design. Please read more about it on google and you can understand how to use it in this case. Its mostly use for impact analysis.

Using the average treatment effect models can sort your problem. As already stated in some earlier posts, propensity score estimators have been widely used to determine the effect of interventions--even without baseline data. For instance, using the nearest neighbor approach (matching estimator), the average treatment effect of the treated can be easily determined. The common problem with observational studies (where one observation can be made at a time) is partially addressed by estimating the counterfactual outcomes. 

I find case studies useful for monitoring change over time if there is no baseline information. Ask the respondents what was your situation before? what changes did you make? what have been the impacts so far? But you have to do this early on in the intervention process so that the first impact change report is akin to a baseline. You can then monitor and measure impacts over time, be they physical, financial, environmental, personal, social, cultural, knowledge based etc,

you can use the focused group discussion methodology and case studies too to do an assessment before and after by indepth interactions and interviews and also looking into the various progress reports during the intervention becomes handy

Thank to all people who have answered this question. 

You could perhaps choose part of the sample from areas unaffected by agricultural extension and the rest from comparable areas where ag ext has been introduced. A comparison of the two should give you some idea about the effects of ag ext.

You use environmental or impact analysis criteria. ISO standard will be better for testing and measurements. Experts will identify the intensity of impacts even baseline data are absent of the particular intervention. Every intervention in the earth is not occur once in a time so that intervention already found in the next place, we can use that reference also.

This can be done by carrying out a baseline survey to assess k'ge before intervention of agricultural extention and another baseline survey after the intervention to assess k'ge after the intervention. Compare k'ge before and k'ge after.

But we should be aware of the issue of spillover. There might be other factors outside the intervention observed or unobserved that brought about the changes.

Several useful suggestions have been made. It is not clear whether you are trying to assess effect of any particular extension project or intervention. This is important because general extension service is diffused everywhere in varying degrees. But within that environment, a partial a new intervention with specific objectives can be made. In that case assessment of impact is quite complex. The attached paper may give yo some idea about possible approach to adopt. In the absence of baseline data, recall data for baseline can be collected as suggested by commentators though usual recall error has to be accepted and carefully handled.

Sometimes statistical tools are used to address data limitations but I do not think that is a proper approach to overcome real data limitations. Complicated statistics can rarely overcome data gaps.

Hope this is useful

Mohammad