Scottish household survey 2018: methodology and fieldwork outcomes

6 Survey Response

Summary

• The final number of social survey interviews achieved was 10,532, a response rate of 64 per cent which matches the target.
• The SHS response rate has fallen in recent years, having fallen by 4% since 1999. The response rate of 64% in 2018 was the same as in the previous three years but 3 percentage points lower than the 2014 response rate of 67%.
• The long-term average response rate for 1999-2011 was 67.9%. However, it should be noted that the calculation has changed slightly for 2012 onwards as a portion of the addresses of unknown eligibility are considered to be eligible whereas previously they would all have been classed as ineligible. This calculation change would have led to a lower response rate in years prior to 2012, if it had been calculated on the same basis.
• There was a target of at least 80 completed physical surveys for each local authority along with a target of 3,004 surveys for Scotland.

6.1 Introduction

This section presents the fieldwork outcomes for the sampled addresses.

The final number of social survey interviews achieved was 10,532. This met the target of 10,450 and represents a response rate of 64 per cent. Since the design changes to the SHS were introduced in 2012 the target number of interviews has been met in recent years, namely 2017 and 2018. However, the SHS response rate has fallen by 4% since 1999 and the impact of this drop in response rate is discussed in section 6.3. The performance of the physical survey is described in section 6.5.

The SHS response rate has fallen in recent years, having fallen by 4% since 1999. The reduction in fieldwork performance resulted in a response rate of 64% in 2018. This was the same as in the previous three years and 3 percentage points lower than the 2014 survey response rate of 67%. The maintenance of the response rate in a climate of declining response rates on other population surveys is commendable.

Previously the contractual required response rate was 67%. However, in 2017, the SHS steering group agreed to lower the target in the new contract. The group agreed on a 65% rate for the contract specification 2018-2021. The 2018-2021 contract states that "the target response rate for the survey will be at least 65%".

The SHS 2018-2021 contract states that "The household response rate assumption for each Local Authority will be updated annually and set as the mean response rate for each Local Authority over the last three sweeps of the survey for which response rate data are available. Response rate here is the successful completion of the household interview." This is subject to the following conditions: (i) the response rate assumption for any Local Authority will not be below 55%; (ii) the response rate assumption for any Local Authority will not be above 80%; and (iii) the Scotland level response rate will not be below 65%. If the third condition is at risk of being breached then each Local Authority's response rate assumption will be uniformly increased to ensure the Scotland level response is 65% or above for each survey year.

Survey response is an important indicator of survey quality as non-response can introduce bias into survey estimates. Standardised outcome codes (based on an updated version of those published in Lynn et al (2001)^[18]) for survey fieldwork were applied across the SHS, SHeS and SCJS. The outcome codes paper includes guidance on the appropriate categorisation of interview outcomes. This will allow consistent reporting of fieldwork performance and effective comparison between the performance of the surveys.

6.2 Scotland-Level Summary

Table 6.1 shows a detailed breakdown of the SHS response for all sampled addresses for Scotland. The addresses of unknown eligibility have been allocated as eligible and ineligible proportional to the levels of eligibility for the remainder of the sample. This approach provides a conservative estimate of the response rate as it estimates a high proportion of eligible cases amongst the unknown eligibility addresses.

The table shows that the overall household response rate was 63.6 per cent. This is below the long-term (1999/2000 to 2011) average response rate for the SHS of 67.9 per cent and below the 2012 to 2014 rates. The effects of the drop in response rate are discussed in section 6.3. It should be noted that from 2012 the calculation was changed slightly and a portion of the addresses of unknown eligibility are now considered to be eligible, whereas previously they would all have been classed as ineligible.

Table 6.1: Fieldwork outcomes, Scotland

Fieldwork Outcome	Sample	% of issued	% of eligible
Responding households	10,532	58.3%	63.6%
Random adult interview complete	9,702	53.7%	58.6%
Refused
Office refusal	472	2.6%	2.9%
Refusal at introduction/before interview	3,459	19.2%	20.9%
Broken appointment - no re-contact	272	1.5%	1.6%
Total refused	4,203	23.3%	25.4%
Non-contact
No contact with anyone at the address	615	3.4%	3.7%
Contact made at address, but not with target respondent	398	2.2%	2.4%
Total non-contact	1,013	5.6%	6.1%
Other non-response
Ill at home during field period	41	0.2%	0.2%
Away or in hospital throughout field period	127	0.7%	0.8%
Physically or mentally unable/incompetent	147	0.8%	0.9%
Language barrier	58	0.3%	0.4%
Lost interview	2	0.0%	0.0%
Other non-response (not covered by categories above)	0	0.0%	0.0%
Total other non-response	375	2.1%	2.3%
Unknown eligibility
Inaccessible	0	0.0%
Unable to locate address	77	0.4%
Unknown if occupied, due to non-contact	365	2.0%
Other unknown eligibility	19	0.1%
Total unknown eligibility	461	2.6%
Estimated eligible addresses in set of unknown eligibility addresses	432	2.4%	2.6%
Total eligible addresses	16,555	91.7%	100.0%
Not eligible
Not yet built / under construction	9	0.0%
Demolished/derelict	56	0.3%
Vacant/empty	861	4.8%
Non-residential	252	1.4%
Address occupied but not resident household	211	1.2%
Communal establishment / institution	48	0.3%
Other ineligible	31	0.2%
Estimated ineligible addresses in set of unknown eligibility addresses	29	0.2%
Total not eligible	1,497	8.3%
All issued addresses	18,052	100.0%

6.3 Drop in response rate

The reduction in fieldwork performance resulted in a response rate of 64% for 2018. This was the same as in previous three years and 3 percentage points lower than the 2014 survey response rate of 67%. Response rates are shown in the Table 6.2 below.

Table 6.2: Response rates over time

	2001/02	2003/04	2005/06	2007/08	2009/10	2011	2012	2013	2014	2015	2016	2017	2018
Response rate	67%	68%	69%	66%	69%	69%	67%	67%	67%	64%	64%	64%	64%
Achieved sample	30,639	30,822	31,013	27,238	28,404	14,358	10,644	10,652	10,633	10,325	10,470	10,683	10,532

The long-term average response rate for 1999-2011 was 67.9%. However, it should be noted that the calculation has changed slightly for 2012 as a portion of the addresses of unknown eligibility are considered to be eligible whereas previously they would all have been classed as ineligible. This calculation change would have led to a lower response rate in years prior to 2012, if it had been calculated on the same basis.

Due to a drop in the response rate in 2015, an analysis was conducted to investigate the likely impact of this drop in response rates. This was undertaken by modeling the 2014 sample results to examine what the impact of a 3% drop in response rates would have been had a lower number of interviews been achieved, and comparing this against the 67% full sample results.

Analysis showed that the demographic composition of the sample was largely unchanged (age, gender), with only the most deprived SIMD quintile and other urban areas showing a 1 percentage point drop in their respective shares of the total sample.

Analysis of mean differences in the population estimates from the two samples for a basket of full sample questions from the SHS, and full one third sample questions from the SHCS module, were also undertaken.

This showed that the absolute mean differences for the total population estimates across the different basket of questions within the household and random adult parts of the survey, including the Scotland Performs National Indicators, were very small, at around 0.1 percentage points.

A few sub-group categories had one or two 'maximum' differences in estimates of around 1 percentage points, including age and social and private-rented sector sub-groups. However, these differences are unlikely to be statistically significant due to small subgroup sample sizes.

Sub-national analysis was not considered. It is expected that there would be a greater impact of this lower response rate for Local Authorities and other sub-national geographies.

6.4 Local authority performance

Table 6.3 shows levels of ineligible addresses, response rate and random adult conversion.

Table 6.3: Response rate and eligibility by local authority

		Ineligible addresses		Responding households		Random adult interviews
Local Authority	Sampled addresses	n	%	n	% of eligible	n	% of HH ints
Aberdeen City	691	99	14%	345	58%	329	95%
Aberdeenshire	622	74	12%	373	68%	361	97%
Angus	435	42	10%	236	60%	205	87%
Argyll and Bute	444	71	16%	244	65%	240	98%
Clackmannanshire	373	13	3%	229	64%	205	90%
Dumfries and Galloway	433	53	12%	260	68%	240	92%
Dundee City	389	20	5%	282	76%	246	87%
East Ayrshire	475	40	8%	250	57%	235	94%
East Dunbartonshire	419	11	3%	275	67%	261	95%
East Lothian	437	35	8%	265	66%	245	92%
East Renfrewshire	466	15	3%	269	60%	245	91%
Edinburgh City	1,469	116	8%	828	61%	778	94%
Falkirk	315	18	6%	225	76%	214	95%
Fife	883	48	5%	591	71%	518	88%
Glasgow City	1,991	114	6%	976	52%	909	93%
Highland	638	68	11%	352	62%	326	93%
Inverclyde	424	42	10%	242	63%	236	98%
Midlothian	479	20	4%	276	60%	254	92%
Moray	424	43	10%	229	60%	219	96%
Na h-Eileanan Siar	403	56	14%	302	87%	298	99%
North Ayrshire	437	44	10%	245	62%	221	90%
North Lanarkshire	875	52	6%	520	63%	444	85%
Orkney	369	54	15%	262	83%	256	98%
Perth and Kinross	432	39	9%	238	61%	217	91%
Renfrewshire	468	30	6%	269	61%	241	90%
Scottish Borders	408	42	10%	257	70%	230	89%
Shetland	367	37	10%	252	76%	227	90%
South Ayrshire	444	28	6%	241	58%	223	93%
South Lanarkshire	866	78	9%	443	56%	384	87%
Stirling	381	30	8%	247	70%	229	93%
West Dunbartonshire	418	28	7%	243	62%	227	93%
West Lothian	377	8	2%	266	72%	239	90%
Scotland	18,052	1,468	8%	10,532	64%	9,702	92%

As in previous years, Argyll and Bute, Dumfries and Galloway, Na h-Eileanan Siar, Highland and Orkney Islands^[19] were among the areas where the highest levels of ineligible addresses were recorded. For Na h-Eileanan Siar and Argyll and Bute, high levels of ineligible addresses were expected as both areas contain a high number of holiday and second homes which are not eligible for the survey. Expected levels of ineligible addresses for Dumfries and Galloway, Highland and Orkney were all above 10%.

The two lowest household response rates in 2018 were in Glasgow City and South Lanarkshire, with Glasgow City having the lowest household response rate. In addition, four other local authorities had a response rate under 60 per cent in 2018.

The conversion from household interview to random adult completion was 92 per cent in 2018 (this was 92% in 2016 and 2017, 91% in 2015, 92% in 2014 and 93% in 2012 and 2013).

6.5 Monitoring and reducing the respondent burden

The Code of Practice for Statistics^[20] states that "Statistics producers should be transparent in their approach to monitoring and reducing the burden on those providing their information, and on those involved in collecting, recording and supplying data." Furthermore it states that "The burden imposed should be proportionate to the benefits arising from the use of the statistics."

The following steps are/have been taken to reduce respondent burden.

• Addresses selected for any of the surveys (SHS, SHeS, SCJS) are removed from the sample frame for a minimum of 4 years so that they cannot be re-sampled for another survey.
• Samples sizes are no greater than required to obtain robust local authority estimates.
• The size of the SHS questionnaire was not allowed to increase following the recent consultation^[21] on its contents. Where there was a strong policy need for new questions, the space/time required for these new questions was created by dropping existing questions or moving them to a biennial basis.
• The SHS social interview is carried out using Computer Aided Personal Interviewing (CAPI). Routing is built into the CAPI script which ensures that respondents are only asked questions that are relevant to them, e.g. respondents are only asked questions on mortgage repayments if they have a mortgage.
• Interviewer briefings are held annually. These events provide interviewers with an opportunity to feedback on improvements that could be made to the questionnaire to assist them conducting interviews and reduce the burden on respondents.
• Interviewers were asked to choose the logo on the tote bag which is being given to respondents as an incentive to complete the survey in 2019. The incentive experiment will run in 2020.
• The Scottish Household Survey team engages its users in the design of new products and in the steering and evaluation of the dissemination of the survey results. The Scottish Household Survey run an annual user day and we normally have around 70-80 attendees every year. We advertise this on the Scotstat email distribution list and through twitter. In addition, we send out annual report evaluation questionnaires to our users to find out views of our publications.
• For large-scale new developments such as the change away from our Excel tables to our new web-based data explorer, we have collected views user through quantitative questionnaires – but also through face-to-face testing and meetings with local authority staff.

6.6 Physical survey

As described in section 0, just under half of the SHS sample was assigned to the physical survey module. For completion of the physical survey, respondents had to agree to make an appointment for a surveyor to make a follow-up visit and to complete the appointment. Table 6.4 shows the number of households assigned to the physical module which responded to the main SHS and the conversion to completion of the physical survey.

There was a target of at least 80 completed physical surveys for each local authority along with a target of 3,004 surveys for Scotland. The result of continuing lower than estimated levels of conversion from household survey to physical survey, meant that there were fewer than 80 physical survey responses for 11 of the local authorities in 2018^[22]. However, the decision to 'open up' an additional stream to the physical survey gave a lower shortfall on the minimum target of 80 per local authority than in previous years (9 local authorities in 2017, 16 in 2016, 17 in 2015, 24 in 2014, 21 in 2013 and 20 in 2012). The 2,971 surveys achieved for Scotland in 2018 was 33 fewer than the target of 3,004.

Table 6.4: Conversion to full physical survey

Local Authority	Conversion rate from household interview to physical survey	Physical survey complete
Aberdeen City	53.1%	79
Aberdeenshire	60.9%	78
Angus	78.7%	96
Argyll and Bute	64.5%	81
Clackmannanshire	64.4%	65
Dumfries and Galloway	70.4%	88
Dundee City	62.7%	89
East Ayrshire	68.9%	84
East Dunbartonshire	59.9%	91
East Lothian	63.7%	80
East Renfrewshire	58.2%	85
Edinburgh City	59.6%	202
Falkirk	70.5%	79
Fife	66.0%	161
Glasgow City	50.5%	204
Highland	56.7%	80
Inverclyde	60.2%	74
Midlothian	61.6%	93
Moray	55.2%	59
Na h-Eileanan Siar	67.9%	95
North Ayrshire	55.3%	69
North Lanarkshire	53.2%	109
Orkney	71.0%	88
Perth and Kinross	71.7%	91
Renfrewshire	61.3%	73
Scottish Borders	53.7%	73
Shetland	81.2%	95
South Ayrshire	57.4%	66
South Lanarkshire	51.8%	99
Stirling	72.7%	88
West Dunbartonshire	57.8%	76
West Lothian	60.4%	81
Scotland	61.4%	2,971