Understanding relational values in cultural landscapes in Romania and Germany

Relational values recently emerged as a concept to comprehensively
understand and communicate the many values of nature. Relational values
can be defined as preferences and principles about human-nature
relationships and focus both on human-nature connections, as well as
human-human connections. Here, drawing on 819 face-to-face questionnaires,
we analysed relational, intrinsic and instrumental values across a total
of six agricultural landscapes in Transylvania (Romania) and Lower Saxony
(Germany). The landscapes described a gradient of land use intensity,
within and across the countries. Our results suggest a bundling of values
into four groups: those concerned with individual cognition (including
intrinsic values), those that focus on nature as a place for social
interaction and relaxation, those that capture cultural identity and
spiritual values and one bundle that only includes instrumental values.
These different values, in turn, were strongly related to (i) respondents’
attitudes towards environmental conservation and the (ii) frequency with
which respondents used nature as a resource.  Instrumental values
have the tendency to be inversely related with relational values and were
found to increase with the land use intensity of the focal landscapes.

Data
collection Preparation for our quantitative survey
included extensive theoretical and literature studies on relational values
and human-nature relationships. Building upon prior empirical work in the
region (see e.g. (Balázsi et al., 2019; Hartel et al., 2016; Riechers et
al., 2019) the questionnaire development included two focus groups with
laypersons to improve structure and wording of the questionnaire and a
pilot study with n = 20. The questionnaire contained
parts on (1) utilisation of nature (Visiting frequency of natural areas in
the vicinity from “daily” to “never”; distance travelled to these places
from “up to 1km” to “over 10km”, use of different natural products such as
water, wood, decorative material from “always” to “never”) (2) attitudes
towards nature and nature conservation (importance from “very important”
to “not important” of the conservation of specific natural attributes in
the landscape), (3) relational, intrinsic and instrumental values and (4)
socio-demographic information (see Supplementary Material S1 for the full
questionnaire). In our study we focused on nine relational values that
were seen as important from our prior research, instrumental and intrinsic
values. An overview of the values used in this paper and their description
can be found in Table 1. Data were collected through face-to-face surveys,
within randomly chosen villages within the focal landscapes. We used
proportionate sampling based on the population density of the villages in
the focal landscapes. Within the villages the streets and households were
sampled randomly. Surveys were conducted on various days of the week.
After a second unsuccessful try, selected households were marked as
dropouts. To decrease the dropout rate we did not randomly select
respondents within a given household. All respondents were asked for an
oral consent to participate in this study, as a personal signature was
deemed to create discomfort and increase drop-out rates, especially in
Romania. Data were collected between April and July 2017. This resulted in
a total sample size of n = 819 across 52 villages
(Romania n = 22, Germany n = 30).
The ethics approval of this research was granted by the Leuphana
University.  
Data
analysis Exploratory factor analysis
Our relational value data frame had
a size of N = 819 observations of 18 variables. We imputed missing data
with the method of predictive mean matching. Cronbach’s α for these
variables was 0.83, while Kaiser-Meyer-Olkin’s measure of sampling
adequacy was 0.93, well above the recommended value of 0.6, Bartlett’s
test of sphericity was significant (χ² (153) = 5583.0, p < .001).
All of these diagnostics suggest reasonable factorability. We considered three, four and five-factor
models using oblimin rotation and a minimum residual factoring method.
Associated scree plots and fit statistics indicated that the four-factor
model was sufficient (RMSEA = 0.071, Tucker-Lewis-Index = 0.885). The four
factors explained 29%, 7%, 5% and 4% of the variance respectively for a
total of 45%. We refrained from removing items with factor loadings
<0.4 because of our sample size of well above 300 (Stevens, 2002
:395). We provide the full loadings matrix in Table 3. We created
composite scores for each factor by adding the scores of the items loading
onto each factor for subsequent regression analysis.   Candidate modeling We modelled the response of the three
latent factors to a set of socio-demographic variables using beta
regression models (Cribari-Nieto and Zeileis, 2010; Grün et al., 2012) on
the latent factor scores that we transformed to the open standard unit
interval (0, 1). The transformation applied was the one recommended by
Smithson and Verkuilen (Smithson and Verkuilen, 2006), so that
y’ = (y × (n – 1)
+ 0.5) / n where y is the data of
length n. We based the set of candidate models on
grouping explanatory variables into three categories: personal
characteristics of the respondent (‘P’: gender, age), nature-based
variables (‘N’: distance travelled, attitude towards conservation,
visiting frequency, frequency of use of natural products) and focal
landscape (‘L’). We constructed the following set of eight candidate
models, which may be seen as our hypotheses regarding what variables might
explain the latent factor scores observed: Null, N, P, L, N + P, N + L, L
+ P, N + P + L. We based model selection on AICc values and used the full
average method where model averaging was required (Grueber et al., 2011;
Nakagawa and Freckleton, 2011). We conducted our analyses using the R
programming language (R Core Team 2019). We present the coefficients of
the best-fitting models for each latent factor in Tables 4 and 5 and in
the supplementary tables A1-A4.

Data has been anonymised.