during the interview. Additional sections may in- 

 clude recreation activities, attitudes, recreation pref- 

 erences, and projected use of proposed new recre- 

 ation facilities. To minimize inconvenience to re- 

 spondents and to avoid respondent fatigue and 

 lapses of concentration, the complete interview 

 should typically not require more than 30 minutes. 



(ii) Pretesting. (A) The basic survey instrument, 

 including bidding formats and questions to collect 

 additional data (e.g., demographic data, respond- 

 ent's history of use of recreation facilities, etc.), 

 must be pretested by using a sample of at least 30 

 respondents in order to generate a data set permit- 

 ting appropriate statistical tests. The pretest sample 

 should not be drawn from the same population as 

 the actual study sample. Sampling procedures for 

 the pretest are not especially crucial, but an at- 

 tempt should be made to obtain a demographic 

 cross section of users. A variety of bidding formats, 

 hypothetical market designs, and payment vehicles 

 should be pretested. 



(B) Nonresponses and protest responses should 

 be tabulated for all bidding formats. Those formats 

 eliciting large proportions (i.e., more than 15 per- 

 cent) of such responses should be eliminated or re- 

 designed and retested. Statistical tests for informa- 

 tion bias, vehicle bias, and starting point bias 

 should be performed, and formats that generate 

 any of these biases should be eliminated, or rede- 

 signed and retested. 



(iii) Sampling. (A) Following pretesting and, if nec- 

 essary, redesign, a sampling frame for the main 

 survey should be drawn. The household is the 

 basic sampling unit. For estimation of activity 

 values, samples may be drawn from reliable lists of 

 participants (e.g., fishing license holders), if availa- 

 ble. For activity values where no such lists exist, 

 and for existence values, the sample must be 

 drawn from the regional population of households. 



(B) Sampling procedures should have the per- 

 formance characteristics of random sampling. To 

 save travel time in a personal interview survey, ran- 

 domized, cluster sampling is permissible, provided 

 that no cluster is larger than one-thirtieth of the 

 sample size. Sample size should be no fewer than 

 200 households. The respondent selected to 

 answer on behalf of the household should prefer- 

 ably be the head-of-household or spouse of the 

 head. In the absence of the head and spouse, an- 

 other adult member of the household may be inter- 

 viewed, provided he or she has assumed a respon- 

 sible life-role (e.g., is a parent or is financially self- 

 supporting). 



(C) Random sampling methods are also used for 

 mail surveys. At least two followup mailings are 

 necessary to reduce nonresponse. In addition, a 

 random telephone survey of 10 percent of the non- 



responses after the second followup mailing is nec- 

 essary. The results of the telephone sun/ey must 

 be analyzed separately in order to permit testing for 

 nonresponse bias. 



(iv) Specific proposed project requirements. (A) 

 Procedures for valuing recreation benefits using 

 project-specific iterative bidding formats are similar, 

 in some respects, to the procedures described 

 above. Aspects that are different are highlighted in 

 the following: 



(B) The population to be sampled is that of the 

 market area(s) for the various categories of recrea- 

 tion opportunities that would be beneficially or ad- 

 versely affected. Survey instruments follow the 

 basic format described above, with the major ex- 

 ception that the bidding formats provide site-specif- 

 ic information on the proposed project itself. Photo- 

 graphs and other stimuli should be focused on the 

 without-project condition for adverse effects and on 

 the with-project condition for beneficial effects. In 

 the latter case, it may be necessary to use photo- 

 graphs of a completed similar project. 



(C) Individual bid data must be used as observa- 

 tions to test carefully for biases, including vehicle 

 bias, information bias, starting point bias, and stra- 

 tegic bias, using established statistical testing pro- 

 cedures. Evidence of bias should (/) lead to elimi- 

 nation of formats producing bias at the pretest 

 stage, and (2) lead to reporting of any bias remain- 

 ing after all instrument redesign possibilities have 

 been exhausted. Final bids are aggregated across 

 the sample and then projected to the market area 

 population. These "population aggregate bids" are 

 then used as estimates of the total value, positive 

 or negative, of the effects, beneficial or adverse, of 

 the proposed increments or decrements in recrea- 

 tion opportunities. Net project recreation effects are 

 calculated as in (e) (1) of this appendix. 



Appendix 3 to Section VIII— Unit Day 

 Value Method 



The unit day value (UDV) method for estimating 

 recreation benefits relies on expert or informed 

 opinion and judgment to approximate the average 

 willingness to pay of users of Federal or Federally 

 assisted recreation resources. If an agency can 

 demonstrate that more reliable TCM or CVM esti- 

 mates are either not feasible or not justified for the 

 particular project under study, as discussed under 

 applicability critena, the UDV method may be used; 

 by applying a carefully thought-out and adjusted 

 unit day value to estimated use, an approximation 

 is obtained that may be used as an estimate of pro- 

 ject recreation benefits. 



83 



