1 Introduction

Transformation towards a sustainable food system for growing urban populations is a major societal challenge (Wiskerke and Viljoen 2012; de Janvry 2019) that makes adaptation essential. Several key challenges of food systems have been identified: current food systems are strongly dependent on global supply chains (Fader et al. 2013), and the food supply of cities is often spatially disconnected from the surrounding agriculture (Billen et al. 2009; Moschitz and Frick 2021). This lack of connectivity between cities and the agricultural production in the cities’ immediate hinterland is increasingly seen as a threat to urban food security (Barthel et al. 2015). In this respect, a regionalization of food supply chains shall promote food system resilience by reducing the impact of disturbances and shocks that can spillover from global supply chains (Lengnick et al. 2015), e.g., the COVID-19 pandemic and the 2022 Russian invasion of Ukraine. Having said that, there are also trade-offs to consider, for example, between diversity of farm practices and resource-use efficiency, as diversity could make food production less efficient (Schlich and Fleissner 2005). Changing food consumption patterns, e.g., towards more regional food, can thus play a critical role in making food systems more sustainable.

It is, however, not straightforward how changing food consumption patterns impact changes in farming practices. Janssen and Van Ittersum (2007), for example, argue that without considering farmers’ responses, the modelled impact of external drivers on farms may be grossly overestimated. This means that elaborated, quantitative information on the drivers of the adaptation decisions is required (Wens et al. 2021), going beyond assumptions about farmers’ decisions as to what to produce and how (Smit et al. 1996). Therefore, including the view of the farmer as an essential land use decision-maker in this process allows to adequately capture the behavioral drivers behind decision-making and broadens the knowledge and perspectives considered (Therond et al. 2009). Since farmers play a key role in shaping the food system, it is vital to investigate the choice of farm adaptations and identify potential barriers.

A growing number of studies is investigating aspects of farmers’ adaptation behavior, for example, adaptation drivers (Dessart et al. 2019), adaptation synergies (Rosenzweig and Tubiello 2007), and transformative adaptation (Käyhkö et al. 2020). Yet, empirical evidence on the relationship between changed food consumption patterns and farmers’ adaptation behavior has received little attention in the literature (see, e.g., Herens et al. 2018; Stringer et al. 2020). A noteworthy study by Hammond et al. (2017) surveyed farmers of different farm types and investigated their motivations and stated intentions to adapt farm practices. However, these studies are restricted to a subset of farm types, lack a distinction between farm types, and do not describe in more detail how adaptation paths may differ across farm types. A research gap was identified in assessing farmers’ adaptation decisions and the conditions under which such adaptations might be employed in the context of a large-scale shift towards more sustainable food consumption patterns. It is argued that this shift towards more sustainable food consumption patterns encompasses, in addition to increasing the share of regional food in diets as described above, increased organic food consumption and decreased meat consumption (Guyomard et al. 2012; Joseph et al. 2019; Rüschhoff et al. 2021; Lauk et al. 2022). While organic agriculture generally produces lower yields than conventional farming, it requires fewer external inputs, promotes balanced nutrient flows, conserves soils, and supports biodiversity conservation (Mäder et al. 2002; Reganold and Wachter 2016). In case of meat consumption, a high share of meat is part of affluent food consumption patterns, which would need to change to supply a growing global population with sufficient food and mitigate greenhouse gas emissions from agriculture (Pradhan et al. 2014; Springmann et al. 2018).

The objective of this study is to assess the adaptation behavior of farmers in response to changed urban food consumption patterns. This is pursued based on the example of the metropolitan region of Austria’s capital Vienna, a major and fast-growing city in the European Union (EU). By means of a set of urban dietary scenarios, far-reaching changes in food consumption patterns in Vienna are introduced in the form of three driving factors: (1) increased consumption of regional food, (2) increased consumption of organic food, and (3) decreased consumption of meat. In a survey, the scenarios were outlined to farmers, and their responses were used to model adaptation paths. The research questions are (1) what adaptation paths do farmers employ in terms of production orientation (crop and livestock production patterns) and production mode (conventional or organic) in light of the demand changes described in the scenarios and (2) what are the main factors and barriers behind the decision to adapt production orientation?

This study advances the field of farm adaptation research in three aspects. First, it offers an assessment determining which of the factors used to develop the dietary scenarios (regional food, organic food, meat) are relevant to farmers’ adaptation decisions. Second, it assesses the adaptability of defined farm types by focusing on switches between farm types. This is relevant as previous research has shown that farm characteristics are an important driver for switching processes (Xu et al. 2020). Furthermore, assessing adaptability reveals the role of path dependency, the strong influence of previous conditions on decision-making, of farmers from different farm types. Third, the study shows the adaptation paths of how different farm types switch in light of the demand changes described in the scenarios in order to better understand the response patterns and heterogeneity of farmers. This allows to better identify key drivers for agricultural policy design towards a more sustainable direction.

The remainder of this study is structured as follows. The following section presents the study region and the design of the survey implemented. In Section 3, we describe farmers’ adaptation behavior. In the last sections, we discuss the results in the context of adaptation paths towards other farm types and draw conclusions on policy measures for a transformative food system change.

2 Methods

2.1 Study region

This study is part of a larger research project that investigates the social, economic, and ecological impacts on the food system in the metropolitan region of Vienna resulting from changed food consumption patterns in Vienna. The study region is the metropolitan region of Vienna, within a radial distance of approximately 100 km around Vienna. The region has been demarcated at the municipality level within Austrian territory (territory of neighboring countries has been excluded, see Fig. 1).

Fig. 1
figure 1

Location of the metropolitan region of Vienna, analyzed in this study, in Austria (left map, upper side), and land use in this region in 2018 (right map); sources: own drawing, based on data from Copernicus Land Monitoring Service (2023) and Statistik Austria (2023)

Full size image

Vienna is situated in the Danube Basin, one of the most important agricultural regions of Austria. The metropolitan region of Vienna offers diverse cropping and livestock systems, including grassland and cattle livestock. The region with an agricultural area of 880,372 ha and 25,696 farms in 2015 (BMNT 2015) contains a large proportion of high-quality, fertile soils, like many other regions in which settlements have historically evolved around the world (Avellan et al. 2012). In total, the agricultural area consists of 81.2% arable land, 13.8% grassland, 4.0% vineyards, 0.3% orchard, and 0.8% others. The primary crops are cereals—in particular wheat, barley, and maize—as well as field forage and sugar beet (BMNT 2015). In 2017, the metropolitan region of Vienna, analyzed in this study, had 3,587,838 inhabitants of which 1,867,582 lived in Vienna (Statistik Austria 2022).

Different approaches exist to define a regional food supply area for a territorial level assessment (Feldmann and Hamm 2015). The setting of the study is the connection between food supply and demand in the metropolitan region of Vienna and how farmers in this region respond to assumed demand changes in Vienna. Therefore, as Clancy and Ruhf (2010) suggested, we based the delimitation of the metropolitan region of Vienna on the geographical extent required to supply the city with food. We compared the regional population and farming structure of different scales between 50 and 150 km around Vienna within Austria. This was done by involving spatially explicit data from the Integrated Administration and Control System (IACS), which describes fields and farms with respect to crop and livestock production (BMNT 2015). The radial distance of 100 km around Vienna was chosen since it best accounts for a high diversity in farming systems and has the potential to provide some portion of Vienna’s food needs.

2.2 Theoretical framework

One way to obtain insights into the likely responses of farmers to external changes is to undertake a survey of their adaptation behavior. Adaptation behavior in the context of this study refers to the personal disposition of the farmer to adjust farm management from a need for change related to dissatisfaction with the current situation to take advantage of opportunities or overcome difficulties (Smit and Wandel 2006; Westley et al. 2013; Xu et al. 2018). The conceptual basis is that stated behavioral intention is the most immediate antecedent of actual behavior in comparison to other measures (Fishbein and Ajzen 1975). The higher the stated behavioral intention to engage in a particular behavior, the higher the probability of showing the particular behavior. Therefore, the stated behavioral intention to adapt or not adapt specific farm practices acts as a cognitive driver or barrier and can be used to predict impacts of scenarios on adaptation behavior (Fujii and Gärling 2003; Bijttebier et al. 2018).

The scenarios used to elicit farmers’ adaptation behavior are so-called contingent scenarios. They aim to assess changes in intended behavior contingent on scenarios by applying the contingent behavior method (Englin and Cameron 1996). This method elicits responses to questions about behavior when changes are proposed relative to the status quo (Kooten 1993), such as “what would you do if the demand for organic food would increase?” These types of questions are asked to make statements of the intended behavior under the conditions of the scenario and provide a flexible way to ex ante judgments (Mitchell and Carson 1989). A key advantage of this method is that it enables researchers to elicit information about scenarios that lie beyond the range of observed levels of experience to construct credible estimates of expected behavior (Eiswerth et al. 2000; Haab and Whitehead 2014). Furthermore, contingent behavior method can mitigate framing effects that are present with monetary valuation and reflects more people’s diverse ways of valuing (Neuteleers and Engelen 2015). When revealed preference data is inadequate to capture behavioral responses, analyzing intended adaptation behavior may be the only way to provide a more comprehensive understanding of decision-making regarding adaptation paths (Whitehead et al. 2013; Mozumder and Vásquez 2018). As the far-reaching demand changes faced by farmers in the scenarios of the present study lie outside those observed in the past, farmers’ previous behavior may be inadequate to rely on. Consequently, the contingent behavior method was used in this study.

Debated issues of the contingent behavior method relate to hypothetical bias (Cummings and Taylor 1999). Because the contingent behavior questions being asked in the hypothetical setting are not binding, answers may be biased. This may result in overstating true preferences, for example, because respondents allocate smaller importance to budget or time constraints. However, survey questions can be framed in a way to mitigate hypothetical bias, for example with cheap talk, which is a short text that informs respondents of hypothetical bias, reminds them to answer the contingent behavior questions as if they were real, or both (Penn and Hu 2018). Despite these criticisms about the accuracy of contingent behavior questions, empirical evidence suggests that contingent behavior questions are a valid tool to predict how people will behave under scenario conditions (Thomson and Tansey 1982; Grijalva et al. 2002).

2.3 Survey design

In order to elicit responses from farmers of the metropolitan region of Vienna towards scenarios, we devised an online survey relying on scenarios developed in the larger research project. Our aim was to formulate a standardized questionnaire (described in Section 2.3.1) providing scenarios that were laid out in a narrative and coherent form to describe fundamentally different prospective situations (Reilly and Willenbockel 2010). The scenarios reflected socio-economic transitions towards more sustainable food consumption patterns increasingly discussed in politics and science. In these scenarios, describing several extreme conditions, farmers can choose to adapt or not. For this purpose, as shown in Table 1, two scenario attributes of urban food consumption patterns in Vienna were combined to create the first two scenarios: RegOrgS (consumption of primary regional food and consumption of exclusively organic food) and RegMeatS (consumption of primary regional food and consumption of two-thirds less meat than nowadays). The other scenarios RegS, OrgS, and MeatS consist of one scenario attribute. These scenarios had a lower weight in the survey and were kept short, and more emphasis was placed here on comparability between the attributes; therefore, they were formulated more generally in terms of an increase or decrease.

Table 1 Overview of the scenarios composed of individual scenario attributes; the scenario abbreviations consist of the pertaining scenario attributes, e.g., RegOrgS (Regional Organic Scenario)
Full size table

Respondents were placed in these scenarios in which only necessary information is provided to keep the scenarios understandable. The contingent behavior questions, here, translated from German to English, read, for example, as follows:

Please imagine the following situation: In the next 10 years, there is going to be a fundamental change in consumption patterns of the Viennese population: The Viennese population consumes primarily regional food and exclusively organic food. Would you change your production orientation?

Since these scenarios relate to future situations, it seemed plausible not to use an exact proportion of regional food demand, as possible food self-sufficiency levels of the study region (corresponding to the increased consumption of regional food) vary with the increase of organic food consumption and the decrease in meat consumption. This involves uncertainty about regional food availability. Another reason for not giving an exact proportion for the scenario attribute of regional food is to avoid reductionism, as the assumption of a constant causal factor operating invariably is problematic for the topic of transformation because empirical evidence has shown that causal processes may alternate (Geels and Schot 2007). The rationale behind these scenario attributes is that some farmers expect economic benefits and development opportunities for their farms as a result of changed demand for agricultural products. Further explanations are provided in Online Resource A.

2.3.1 Questionnaire description

The questionnaire was pre-tested by farmers and other local stakeholders of the study region and was adjusted to the feedback received. The final questionnaire consisted of three parts and 31 questions and was implemented with the survey web application LimeSurvey using a responsive web design (see Online resource A for the questions of the questionnaire). The time to complete the questionnaire was about 10 min.

The first part of the questionnaire provided background information on the study and then queried basic farm characteristics, including farm type, details of which are given in the following subsection. The second part of the questionnaire covered the main topic of the questionnaire “whether and what to adapt?” Respondents were presented with a scenario introduction including a cheap talk paragraph to mitigate hypothetical bias. To make the impacts of the scenarios easily understandable to respondents, a note explained that the scenarios affect the supply and demand of agricultural products. The survey presented five separate scenarios in the order given in Table 1Footnote 1, in which the respondent indicated what adaptations, if any, they would make in light of these scenarios. It is important to note that among the various adaptation dimensions farmers have (Smit and Skinner 2002), we consider two adaptation dimensions here: change in production orientation (operational focus with respect to crop and livestock production patterns) and change in production mode (conventional or organic mode of production). Therefore, the answer options available to the respondents were as follows: do not adapt and carry on as before, adapt production mode, adapt production orientation, and adapt both production mode and production orientationFootnote 2. The two scenarios RegOrgS and RegMeatS, each combining two scenario attributes, were the main focus of the survey and queried changes in production orientation and production mode, while the rest was restricted to changes of production orientation and kept far shorter. In RegOrgS and RegMeatS, follow-up questions were asked to explore the further path of respondents’ farm adaptation: If respondents decided to adapt in some way, what specific crop and livestock practices would they change and to what degree, and if they indicated to carry on as before, what are the barriers to adaptation. These questions aimed to elicit crop and livestock practices in the scenarios compared to the current state. The third part of the questionnaire involved questions about sociodemographic characteristics and current farm practices.

2.3.2 Layout of farm types

The existing farms in the metropolitan region of Vienna were classified according to a typology into farm types, to which the respondents in the survey then assigned themselves. This farm typology was used to segment respondents in the survey. Each farm type represents a group of individual farms in the study region that is relatively homogenous in terms of production orientation and production mode. For the purpose of the classification, IACS provided data on the main characteristics of farms (BMNT 2015). Farms were divided into farm types according to three levels of increasing detail. In the first level, the farms were divided according to their production mode (conventional or organic). In the second level, they were further divided according to the “general type of farming” (GTF) of the European community typological classification, which assigns to each farm a prevailing production orientation as described in Commission Delegated Regulation (EU) 1198/2014. The main criterion for this division is the average monetary value of the agricultural output at the farm-gate price, coming from different farm practices (Eurostat 2022). In the third level, GTF were further divided based on mix of crop, mix of livestock, and share of crop production marketed or used as feed on farm, using a rule-based classification of similarly structured farms and hierarchical cluster analysis. This decision was taken to keep the total number of farm types manageable while considering the heterogeneity of farming across the study region. Details of the farm-type classification are given in Online Resource A. The farm types defined are listed for the sake of simplicity without specifying production mode in Table 2 (a full table is provided in the Online Resource A).

Table 2 Overview of farm types laid out by the typology (note that the differentiation between conventional and organic production mode is not shown)
Full size table

2.4 Modelling farmers’ responses

Based on survey responses, we assessed the adaptation decision with a logit model and modelled farm type switches by assigning preferential weights to the stated farm adaptations. Calculations were conducted with Python (packages: pandas, statsmodels, and NumPy) and IBM SPSS.

2.4.1 Logit model

The most common framework to estimate the probability to adapt production orientation as a function of explanatory variables is with a binary logit model. The binary logit model has been widely adopted because it has analytical advantages in dealing with discrete binary outcomes (Cramer 2003). It allows to examine farmer’s adaptation decisions if adaptation is operationalized as a binary choice and to determine the associated probabilities for the choice of a particular adaptation decision (Mabe et al. 2014). An alternative to the binary logit model used in this study would have been to use the random forest method, but a binary logit model was preferred here, because it provides a measure of how relevant an explanatory variable is and gives the direction of association, i.e., positive or negative.

The binary logit model has the general form (Greene 2003):

$${P}_i\left({Y}_i=1\right)=\frac{\exp \left( X\beta \right)}{1+\exp \left( X\beta \right)}$$
(1)

where Pi is the probability of adaptation of production orientation in RegOrgS and RegMeatS, Yi is the dependent variable (Yi = 1 if farmer i adapts production orientation and Yi = 0 otherwise), X is a vector of explanatory variables, and β is a vector of coefficients. To estimate the magnitude of single scenario attributes and the variables described in Table 3 on the probability to adapt production orientation in response to the scenarios in which two scenario attributes were combined (RegOrgS and RegMeatS), the survey generated data on the explanatory variables.

Table 3 Main characteristics of the sample (survey respondents) and the population (farms in the study region)
Full size table

As the estimated logit parameter estimates are not easily interpretable, we calculated marginal effects to quantify the effect sizes of explanatory variables on the probability to change production orientation. To be more specific, we calculated the marginal effect at the means, which measures the change in Pi(Yi = 1) by a 1-unit increase in an explanatory variable of interest, holding all other explanatory variables at their mean values. This means for categorical explanatory variables (which are all dummy-coded and can therefore only take two values), a discrete change from 0 to 1, i.e., changing from off to on, and for continuous explanatory variables, an increase by 1 unit. The sign and statistical significance of the effect exerted by each explanatory variable on the likelihood of adaptation of production orientation are reported in the columns “Parameter estimate” and “P > |z|” of Table 4. The smaller the “P > |z|” value, the higher the probability that the variable in question exerts an effect on the dependent variable (Bertoni and Cavicchioli 2016).

Table 4 Logit results of adaptation of production orientation for RegOrgS and RegMeatS
Full size table

2.4.2 Identifying farm-type adaptations

Modelling farm-type adaptation implied several steps in order to obtain a quantitative assessment of likely responses to the investigated scenarios. For this purpose, responses of the survey were analyzed with a switch index specifically suited to derive farm-type specific preferential weights for adaptation decisions. First, the respondents who indicated to adapt production orientation were selected. Next, responses to the follow-up questions pertaining adaptations of farm practices as a result of adapting production orientation were split according to respondents’ current farm type. This was done to integrate the responses in the scenarios regarding adaptations of farm practices into respondents’ current farm type. The answers regarding adaptations of farm practices, i.e., crop and livestock expansion, indicate the direction of farmers’ adaptation behavior.

For each of the two answer options to adapt production orientation (either only production orientation or production orientation and production mode), the number of respondents who switch to another farm practice was calculated as follows, based on Mehdi et al. (2018) and Sattler and Nagel (2010):

$${s}_{it}={\sum}_{it}\frac{e_{ijt}}{a_{jt}}$$
(2)

where sit is the total share of expansion of each farm practice (i) per farm type (t), eijt is the expansion of the farm practice (i) of farm (j) of farm type (t), and ajt is the total agricultural area of farm (j) of farm type (t). If \(\frac{e_{ijt}}{a_{jt}}\) > 1, it was trimmed to 1 so that all \(\frac{e_{ijt}}{a_{jt}}\) had an identical range between 0 and 1, because otherwise, a farmer would expand more of a crop or livestock category than his total agricultural area or total livestock units, which was not allowed in the survey as explained as part of the pertaining survey question. Next, we calculated dit, the share of expansion of farm practice (i) of farm type (t) in relation to the expansion of all farm practices per farm type by taking the six largest sit to capture the major patterns in the data:

$${d}_{it}=\frac{s_{it}}{\sum_{it}{s}_{it}}$$
(3)

This normalization procedure was used to set the sum of all dit to 1 for each farm type. It follows:

$${n}_{it}={d}_{it}\ast {n}_t$$
(4)

where nit is the number of respondents belonging to farm type (t) who switch to farm practice (i) and nt is the number of respondents of farm type (t) who answered to adapt either only production orientation or production orientation and production mode.

We used nit for the decision rule to select farm types that are most similar to the stated switches of farm practices. This integration of farmers adaptation behavior was done to reflect the stated farm adaptations and to switch the farm from its initial farm type to the adapted farm type of the scenario. In this process, farms switched to already existing farm types because further differentiation of farm types would not bring advantages due to the high number of defined farm types. The switching process implied that adapting production orientation led to a switch of farm type, which can occur in two forms: switch of farm type without changing production mode or switch of farm type with changing production mode (from conventional to organic). Based on these switches between farm types (tmn), the adaptation patterns can be collected in a transition matrix (Q) (Zimmermann et al. 2009):

$$Q=\left[\begin{array}{cccc}{t}_{11}& {t}_{12}& \cdots & {t}_{1n}\\ {}{t}_{21}& {t}_{22}& \cdots & {t}_{2n}\\ {}\vdots & \vdots & \ddots & \vdots \\ {}{t}_{m1}& {t}_{m2}& \cdots & {t}_{mn}\end{array}\right]$$
(5)

and thereupon represented as alluvial diagram, which is shown in the next section.

3 Results

The survey was conducted online in February and early March 2019. The data collection was arranged in cooperation with the agricultural chambers of two Austrian federal states that are located in the study region—Lower Austria and Vienna—and the Federal Ministry for Sustainability and Tourism to distribute the questionnaire. In total, we received 1798 completed responses. Respondents who indicated that their farm type was among those that did not produce primarily food, i.e., forestry and horse-keeping, or that their farm was located outside the study region were dropped from the sample, leaving 1720 completed responses. Since not all of the respondents answered every question, we report the number (n) of respondents for the particular question.

The main characteristics of the sample relative to the population of the farms in the study region are reported in Table 3 and indicate that a wide range of farmers responded to the questionnaire. Although the sample is not representative with respect to GTF, only mixed farming (23% in sample and 14% in population) and specialist grazing livestock (19% and 28%, respectively) differed relatively between the sample and the population. The most dominant GTF in terms of the number of farms and agricultural area managed was specialist field crops, which represented 40% of the sample and also 40% of the population.

The adaptation patterns of respondents to the scenarios were split by scenario and production mode and are presented in Fig. 2. Totalled across RegOrgS and RegMeatS, 41% of respondents would adapt at least their production orientation, and 48% would not adapt and carry on as before. The remainder of the respondents would solely adapt production mode to organic or chose “No answer”. The differences between RegOrgS and RegMeatS concerning production orientation were striking: A high proportion of respondents would adapt at least production orientation in RegOrgS (in total 44% of conventional farms and 53% of organic farms) compared to the proportions in RegMeatS (in total 34% and 38%, respectively). Among respondents of conventional farms in RegS, OrgS, and MeatS, increasing regional food consumption (RegS) induced the highest rate of adaptation of production orientation at 68% (shown in Fig. 2c). Increasing organic food consumption (OrgS) induced 82% of respondents of organic farms to adapt production orientation. RegMeatS and MeatS, both with decreased meat consumption, showed lowest rates of adaptation of at least production mode or production orientation (in total 42% and 32%, respectively). In contrast, scenarios with increased regional food consumption, RegOrgS and RegS, showed highest rates of adaptation (in total 62% and 67%, respectively).

Fig. 2
figure 2

Responses to the survey question of whether to change production orientation and production mode of conventional farmers (a) and organic farmers (b); responses to the survey question of whether to change production orientation of conventional farmers (c) and organic farmers (d). For consistency, “No answer” responses were transformed to the most conservative response, i.e., “No” (in total: n = 1720)

Full size image

Table 4 presents the results of two estimated logit models of adaptation of production orientation for the RegOrgS and RegMeatS scenarios. Since the survey focused mainly on these two scenarios (and less on RegS, OrgS, and MeatS), the logit models relate to these two scenarios. Since specific scenarios encompass specific switching potentials for certain farm types and the overall proportion of observations correctly classified was higher with separate models, we used one logit model for RegOrgS and one for RegMeatS. The model correctly classified 72% of adaptation choices in RegOrgS and 78% in RegMeatS. In RegOrgS, the scenario attribute of increased regional food had the most impact on the probability to adapt production orientation. Specifically, the marginal effect suggested a 0.449 (or to put it another way, 44.9%) increase in the probability of adapting production orientation when respondents’ decision to adapt production orientation in case of increasing regional food consumption changes from 0, i.e., absent, to 1, i.e., present (see line “Adaptation of production orientation in RegS” in Table 4). Respondents’ decision to adapt production orientation in case of increasing organic food consumption had a low marginal effect with 0.086 in RegOrgS, meaning increasing organic food consumption had a low impact on adapting production orientation. A further point is that GTF of respondents plays a significant role in RegOrgS, and education is positively associated with the probability of adapting production orientation and, in most cases, significant in RegMeatS.

Another question of the survey asked “Which food consumption patterns of the Viennese population would be most beneficial to your current farm?” in order to assess the investigated scenario attributes: 70% prefer primarily regional food, 26% prefer exclusively organic food, and 4% prefer two-thirds less meat. This indicated that overall, the scenario attribute “primarily regional food” received the highest popularity among respondents both in the logit model as well as in the question just presented.

The survey provided answers on how respondents would adapt their farm type in the scenarios. These answers were modelled as described in Section 2.4, and the results are shown in Fig. 3 as alluvial diagram (Yeung 2018). The alluvial diagram visualizes switches between GTF at two time points in a network composition. The switches of GTF between the current GTF and the GTF respondents would switch to in the scenario are represented by flows. The size of the flows represents the proportion of respondents per GTF. Because transformative adaptation is essentially grounded on switches between GTF, respondents who solely adapt production mode from conventional to organic farming and respondents who switch within their current GTF were not shown in the alluvial diagram. To avoid over-cluttering, flows that represent less than 3 respondents were excluded. For more detailed switching patterns on a farm-type level instead of a GTF level, some diagrams are provided in Online Resource B.

Fig. 3
figure 3

Alluvial diagrams of RegOrgS (a) and RegMeatS (b) show the adaptation behavior of respondents from the current GTF to the GTF to which respondents would switch in the scenario, disregarding switches in production mode. Flows represent a minimum of 3 respondents per GTF to avoid over-cluttering (nRegOrgS = 189; nRegMeatS = 121)

Full size image

An examination of the switches between farm types revealed that farm types switched mainly within their GTF. In other words, among respondents of the different GTF who responded to switch production orientation, most switches occurred within GTF, representing incremental adaptations (approx. 64%), and less occurred between GTF, representing transformative adaptationsFootnote 3 (approx. 37%). This is identical in RegOrgS and RegMeatS. In particular, respondents of specialist field crops showed this kind of switching pattern—roughly 80% of those who switched, switched within GTF. In contrast, predominantly respondents of mixed farming switched between GTF (84% in RegOrgS and 66% in RegMeatS). This is plausible because farm types of this GTF had more favorable initial conditions to adapt between GTF than farm types of other GTF.

By comparing the alluvial diagrams for RegOrgS and RegMeatS (Fig. 3) several points stand out. The switches in RegOrgS showed a higher diversity than in RegMeatS, resulting in more flows drawn between the GTF. Accordingly, respondents adapted production orientation—considering switches within and between GTF—RegOrgS 1.31 times more frequently than in RegMeatS. Fewer respondents switched to a GTF with livestock in RegMeatS than in RegOrgS because reduced demand for meat implies production declines for GTF with livestock. Instead, the GTF of specialist field crops becomes the predominant GTF to which respondents of other GTF would switch to in RegMeatS, which becomes most apparent for respondents of mixed farming.

Respondents who stated that they would not adapt production orientation in RegOrgS, RegMeatS, or both were asked for their reasons. Figure 4 shows these reasons split by GTF. In total, across all GTF, the most common reason for intending not to adapt production orientation concerned the success with the current farm type, which was given by 48% of the respondents. Among these, respondents of specialist horticulture declared confidence in the current farm type remarkable frequently with 68%. Respondents of specialist granivores chose most frequently “high investment required” with 42%, whereby among respondents of specialist granivores, 34 respondents were from pig farms and 16 from poultry farms. Respondents of specialist permanent crops indicated least frequently the answer option “high investment required” with 16% (153 respondents were from vineyards and 18 from orchards).

Fig. 4
figure 4

Reasons given for why not to adapt production orientation grouped by GTF. Percentage represents the proportion of respondents who gave that reasons; respondents were able to make multiple responses (maximal three) that applied to them (in total: n = 1232)

Full size image

4 Discussion

Our study traced empirically grounded adaptation paths of farmers facing different scenarios of changed urban dietary patterns for a specific metropolitan region of a European country. For most GTF in the metropolitan region of Vienna—especially the most common GTF, namely, specialist field crops—farmer proportions are consistent in sample and population.

Comparing the numbers of adaptations regarding production orientation between organic and conventional farmers reveals that organic farmers are more likely to adapt. This result corresponds to the finding by Weltin et al. (2017) that organic farmers are more likely to diversify farm practices than other farm types. It has also been argued that organic farmers are more familiar with systemic thinking and diverse crop rotations (e.g., knowledge on the functioning of the farm ecosystem), which are fundamental in organic farming (Darnhofer et al. 2010b) and can be beneficial for farm adaptations.

To answer the first research question (what adaptation paths do farmers employ?), the results show that farmers have to some extent a contextually based adaptability to changed food consumption patterns. Clearly, adaptation decisions are constrained by local biophysical conditions for agricultural production, e.g., soil and climate conditions. Adaptation paths that require high investments and institutional involvement, e.g., stable building, are less likely, because the costs of adaptation, including transactions costs, might be higher than the potential benefits (Liebowitz and Margolis 1995). Adaptation decisions are to a high extent path-dependent. Economic profitability as well as the possible future adaptations is closely linked to the current farm type due to a high amount of fixed assets (Andersen et al. 2006). These results are consistent with economic theory on path dependency (Liebowitz and Margolis 1995). If there are uncertain returns, farmers tend to wait and see, because they have little room for maneuver for experiments that require financial and labor resources. As a result, it is more likely to increase existing farming practices or those closely related than to introduce farming practices associated with other GTF. In other words, changes in urban food consumption patterns tend to stimulate piecemeal, incremental adaptations towards existing or well-known farm practices, owing to path dependency and the fact that incremental adaptations appear, initially at least, reassuring (Filho et al. 2022). This makes it challenging to promote transformative farm adaptations (Cradock-Henry et al. 2020) merely by changing food consumption patterns without supporting policy measures (Hammond et al. 2017). To address this concern, farmers could be encouraged to stepwisely concatenate incremental adaptations in farming practices—i.e., low-regret strategies—to lay the groundwork for more profound transformative adaptations as changed food consumption patterns become more prevalent (Cornwell et al. 2021). Once the groundwork is laid, eventual transformative adaptations become also less dramatic (Sutherland et al. 2012), since employed adaptations may reinforce further adaptations as farmers explore new configurations between old and new farm practices and learn more about the employed adaptation direction (Geels and Schot 2007). This may create a development towards considerable adaptations over time either at farm level or sectoral level eventually accumulating to the stage of transformative adaptations—this can be viewed as a co-evolutionary interpretation of adaptation (Rickards and Howden 2012).

In general, more diversified GTF with a broader operational focus show a higher likeliness to adapt. In this light, a high diversity within the farm ensures adaptability (Darnhofer et al. 2010a). As shown in the results, the GTF of mixed farming can adapt livestock or crop production according to future needs easier than other GTF with livestock. This underlines that less specialized farm types have less difficulties coping with external changes and shocks in farming conditions (Nainggolan et al. 2013; Weltin et al. 2017).

Addressing the second research question (what are the main factors and barriers behind the decision to adapt production orientation?), the marginal effect at the means of the logit model indicates a strong, positive effect of the scenario attribute of increased regional food consumption on the decision to adapt production orientation. Together with respondents’ preference towards increased regional food consumption and the higher adaptation rate in RegOrgS compared to RegMeatS, this suggests that farmers adapt rather due to expected advantages than disadvantages avoidance. In Austria, regional food consumption has increased in recent decades, generating attention in the media (Schermer 2015), and consumers are willing to pay a price premium for regional food (Enthoven and Van den Broeck 2021). The social recognition and acknowledgement farmers receive from consumers support their professional self-conception and motivate them to market their products as regional. This encourages farmers to see scenarios with increased regional food consumption as a motivating factor to adapt. Additionally, many farmers formed an emotional attachment with their land and region (Gosling and Williams 2010), meaning that increasing regional food consumption corresponds to their values. To facilitate further development towards regionalization, for example, direct contact among farmers, processors, and consumers (Zasada 2011; Stringer et al. 2020) or an umbrella brand for regional food could be established (Doernberg et al. 2019).

There is some indication that the more specific the scenario attributes are for the farm development, the lower the preference for them. All farmers in the region would benefit from increased regional food consumption without the necessity to adapt into specific directions. By contrast, only those already farming organically or can imagine a conversion to organic farming would benefit from increased organic food consumption. Lastly, for livestock farmers and farmers focusing on feed production, decreased meat consumption would imply price and production declines as well as higher pressure to adapt. Apart from that, there is a general reluctance to decrease meat consumption (Macdiarmid et al. 2016), and, for this reason, respondents might not have believed this scenario attribute to be reasonably feasible. This can cause probability of provision bias (Carson and Cameron 1995), i.e., that instead of respondents assuming decreased meat consumption occurs certainly in the scenarios, they assume that decreased meat consumption is less likely to occur and therefore undervalue this scenario attribute.

The reasons given for not changing production orientation show that maintaining income and success of the current farm type is crucial. The reason for maintaining income is related to the present bias, i.e., to give more importance to immediate costs than to long-term benefits (Doyle 2013), for example, with long-time frames for investment. The reason “successful with the current farm type” shows that farmers who perceive the current situation as satisfactory do not intend to make any transformative changes (Öhlmér et al. 1998). Additionally, this reason may be partially attributed to status quo bias (Samuelson and Zeckhauser 1988). Status quo bias reflects the tendency of avoiding larger changes and selecting a well-known or already employed alternative disproportionally often because of less familiarity with other alternatives or loss-aversion. These results are consistent with findings from previous studies (Rodriguez et al. 2009; Niskanen et al. 2021; Nyberg et al. 2021).

If farmers’ current revenue is set as a reference point, then people weigh losses heavier than gains of the same value (Kahneman and Tversky 1979), making them averse to adaptation. Correspondingly, farmers with less revenue are less likely to adapt (Trujillo-Barrera et al. 2016). In accordance with our results, Dessart et al. (2019) found that being moved by economic objectives makes farmers reluctant to adapt. By contrast, active engagement, openness to new experiences, and environmental awareness were found to be associated with a higher likelihood to adapt. Thus, informing and showing positive perceptions towards the marketability of new products are viable for farmers who are considering adaptation, to make an educated decision for their particular circumstances (Padel et al. 1999; Käyhkö et al. 2020).

Responses of the farmers of specialist granivores suggest that the decision to adapt production orientation is largely based on balancing the economic costs against the economic benefits. This finding further supports the idea of valuing adaptations that yield improvements in profitability (Trujillo-Barrera et al. 2016; Sánchez et al. 2016). For this reason, specialized or intensive farms that are greatly successful on the international market might not find it financially beneficial to adapt only to meet changed Viennese dietary patterns. Also, when a high investment is required, there is an option value to wait with adapting production orientation, as it takes time to assess changes in food consumption patterns and their implications for the farm (Tranter et al. 2007; Musshoff and Hirschauer 2008). A further issue here is that a part of farmers responded not to adapt are locked into their farm type or refuse to consider specific adaptation opportunities owing to a high conviction or caution towards specific farm types and the potential disadvantages, e.g., loss of social position, associated with employing unsuccessful adaptations (Sutherland et al. 2012).

There are policy components that would be required for transformative farm adaptations. Changed food consumption patterns are an important driver for a transformation to a more sustainable food system. However, this study also highlights the need for accompanying policy measures on the production side to encourage transformative farm adaptations. This applies especially to the reduction in consumption of livestock products, which is essential to ensure that sustainable farm practices are prioritized; otherwise, it will be challenging for farmers to choose effective transformative adaptations (Rosenzweig and Tubiello 2007). Policy measures could, for example, include making payments to farmers in the initial phases of adaptation or equipping them with requisite knowledge on farm adaptation to reduce start-up costs and barriers, especially for new farmers (Baumgart-Getz et al. 2012; Sánchez et al. 2016; Arzeni et al. 2021). Advisory services and facilitation of knowledge sharing could foster conditions that empower farmers in adaptation-related decision-making processes for a policy efficient design. Policy design needs to consider the varying susceptibility to adaptation per farm type. Providing farm type specific incentives for adaptation decisions and mitigating structural barriers to adaptation, especially for farms with a high amount of fixed assets, e.g., intensive livestock farms, facilitates sustainable farm trajectories. Although, it is debatable whether policies should support farm types that require low adaptation-related investments or farm types that require high adaptation-related investments. On the one hand, supporting farm types that require low adaptation-related investment would achieve a wider adaptation impulse, in terms of number of farms, with fewer financial resources. On the other hand, farm types that require high adaptation-related investments and do not have sufficient financial capital to bear the adaptation would need financial support to maintain farm income during an adaptation period (MacRae et al. 1990) in order to offer a perspective for farm development and mitigate path dependency. This is also important given that the current CAP provides few resources to adaptability and transformability.

Some limitations of this study should be noted. The adaptation rates in the scenarios are obtained from contingent behavior questions. Farmers tend to make relatively thoughtful decisions when it comes to long-term adaptations (Dessart et al. 2019). Still, due to hypothetical bias in the scenarios and simplistic assumptions about adaptation (Wreford and Adger 2010), the adaptation rates might be overstated. However, the triangulation approach between contingent scenarios and logit model facilitates greater completeness and generalizability of the findings. A limitation to make an appropriate assessment of farmers’ adaptability in response to scenarios towards regional food consumption is that the effects of scenarios towards regional food consumption would need to be set into perspective to farmers’ level of integration into international, agricultural commodity markets, i.e., the extent to which farmers are involved in regional or international supply chains. Farmers who produce predominantly for international, agricultural commodity markets give less weight to scenarios towards regional food consumption for potential farm adaptation than to scenarios that outline demand changes on larger geographical scales. In reality, there are many more possible farm types and adaptation possibilities than we can cover here, but focusing on a farm typology allowed us to highlight the most common adaptation paths. Further limitations are the tendency that dissatisfied people are more willing to adapt as well as that feedback to act in relation to societal norms and spatial dependence is not considered. Moreover, the behavioral intentions captured here are snap-shots of farmers’ decision-making, and there is a need for data on more gradual change over time. As implemented here, only the metropolitan region of Vienna within domestic boundaries is studied. The study region could be expanded to include a transnationalization, as the European integration process is increasingly causing regions to transcend national borders in economic terms. For example, extending the present study region to the so-called Centrope region could be a valuable avenue. The Centrope region consists of a number of adjacent territories along the borders between Austria, Czech Republic, Slovak Republic, and Hungary (Planning Association East 2023).

5 Conclusions

The present study captures farmers likely adaptation paths in response to changed urban food consumption patterns. The findings gained through this study contribute to a better understanding of the behavioral characteristics of farm adaptations for a broad range of farm types and diverse decision-making. Furthermore, the study offers a useful adaptation planning tool to design policies that underpin and route farm adaptations in a sustainable direction. One of the major findings of this study is that each GTF is linked to specific patterns of adaptation paths and barriers to adaptation. In general, less specialized farms have wider options to adapt and are more likely to employ transformative adaptations than more specialized farms, owing to less path dependency.

The evidence from this study suggests that farmers’ transformative adaptations in terms of production orientation, in response to the studied scenarios, are rather restrained. In reality, transformative adaptations fall short of its potential (Mandryk et al. 2014); however, the concatenation of incremental adaptations over longer time scales may eventually lead to a stage of transformative adaptations if sustaining a long-term commitment. Even though transformative adaptations tend to be less common, they are highly relevant as fundamental change of the main focus of production of a farm allows for a greater range of farm adaptations. We conclude that transformative adaptations should be better accounted for in the adjacent field of land-use modelling, as it typically lacks considerations of farmers’ intentions. Assumptions underlying agricultural land-use dynamics that do not involve farmers’ intentions are suitable for incremental adaptations but are inadequate under scenarios that entail transformative changes in existing patterns, which lie beyond the range of observed levels of experience (Renner et al. 2020).

The study’s findings also highlight scenario attributes outlined to respondents in this study: Regionalization of the Viennese food system widely resonated with respondents, showing that cities can be a starting point to reshape the food flows towards regional sources in the cities’ hinterlands. In contrast, when asked which scenario attribute represented the most beneficial scenario attribute for the respondents’ current farm, decreased meat consumption was by far the least frequently chosen. This is a critical issue as incremental adaptations of livestock farms may not be enough to reconfigure food systems towards sustainability. A transformative adaptation from livestock farms with a high amount of fixed assets to other production orientations takes considerable investment and time. In the context of policy design, a transformation of the food system that is socially and economically acceptable for farmers cannot be solved with relying on demand changes alone because of the cross-sectional constitution of food systems, e.g., with the interplay of different agricultural input sectors, processors, distribution channels, and retail structures. In fact, adaptation challenges and implications for farmers also need to be tackled, as farmers are essential actors shaping agricultural landscapes, embedded in the agricultural and food industry, and can be part of the solution to environmental problems. Consequentially, a transformation would require a deep policy integration between coherent, multidimensional production and consumption-oriented policy measures to stimulate transformative adaptation in agriculture (Frison and Clément 2020) that transcends the changes of the investigated scenarios.

Trigger events, in which farmers realize that a change in farm trajectories is needed, are preconditions for adaptation—they include especially financial imperatives as well as farm succession and create opportunities for restructuring the farm (Inwood and Sharp 2012; Sutherland et al. 2012). Policy measures addressing these findings could comprise assisting farmers in the phases of farm succession in enlarging their portfolio of possible adaptation paths that work well for a wide range of circumstances. Another policy measure could be the establishment of a multi-actor knowledge network, available to farmers, that elaborates plural adaptation paths depending on the operational focus to deal with adaptation constraints and path dependency (Šūmane et al. 2018). Furthermore, policy measures aimed at safeguarding metropolitan farms that are connected to the city would need to mitigate loss of agricultural land driven by the continuing trend of urban expansion and subsequent land sealing (Avellan et al. 2012). This is especially relevant for the metropolitan region of Vienna: Being the fifth largest city in the EU, it has seen the highest population growth among the ten largest cities in the EU over the periods 2010–2020 (Statistics Vienna 2022).

Future research may explore the same phenomenon of adaptation by integrating a Markov chain, which requires data of switches from one farm type to another, to involve subjective probability distributions and better account for a temporal dimension in the decision-making process. Data on intended adaptation paths, such as those set out here, also lay the foundation to build land-use models, e.g., agent-based models, parametrized with empirical data on intended adaptation behavior to allow a variation in decision-making apart from purely economic considerations. Furthermore, future research would need to allow for more dynamics to understand how farmers consider adapting when social norms and feedback mechanisms with regards to other farmers, e.g., information diffusion, are included as farm adaptation is not a linear process. Also, investigating farm adaptation by taking into account entrances to farming, stays, and exits out of farming of various farm types using data on actual observed farmer behavior would be a fruitful area for further investigation.