Though rarely stated explicitly, an important rationale for wanting to maximise response rate is an assumption that this will bring greater gains in accuracy of estimation than simply increasing the selected sample size. In other words, it is assumed that adding in hard-to-get respondents will not merely improve precision, but will reduce non-response bias. In this paper, we examine two distinct components of the difficulty of achieving an interview: difficulty of contacting sample members ("ease of contact") and difficulty of obtaining co-operation once contact is made ("reluctance to co-operate"). We assess the separate and combined effects of ease of contact and reluctance on nonresponse bias in order to estimate the impact of extended interviewer efforts on survey error and to answer subsidiary questions about the relative effectiveness of different types of extended efforts. Data come from the Family Resources Survey, Health Survey for England and British Social Attitudes Surveys.