Implementability of SC-Compatible Policies

Abstract

Having identified a list of SC-compatible policies, and given the fact that the preferences of micro and small enterprises (MSEs) are not always the same as those of the medium enterprises (MEs), let alone with the social planners’ perspectives, some of the proposed policies can neither be aligned with the desired goals nor accepted by all agents. Which among those policies and policy-mix that can be aligned? Is there an institutional setting or a mechanism that could be designed to implement those policies? The discussions in this chapter address these questions. The results show that having a network is viewed by MSMEs as a vital step for improving their productivity because through a network they gain more interactions and better linkages among themselves and with other stakeholders.

Having identified a list of SC-compatible policies, and given the fact that the preferences of micro and small enterprises (MSEs) are not always the same as those of the medium enterprises (MEs), let alone with the social planners’ perspectives, some of the proposed policies can neither be aligned with the desired goals nor accepted by all agents. Which among those policies and policy-mix that can be aligned? Is there an institutional setting or a mechanism that could be designed to implement those policies? The discussions in this chapter address these questions. The emphasis is on finding the institutional setting or mechanism that would ensure the selected policies and policy-mix are aligned with the desired goals and accepted by all agents (MSEs and MEs). One of the challenges in trying to identify such a mechanism is the possibility that the information or the policy ranking conveyed by each of the agents may not represent the true conditions due to either the circumstances they were under, or they simply were not being fully truthful when expressing their preferences. This and other issues surrounding the mechanism to implement the selected policies are also discussed.

Unlike in the previous chapter, where our focus was trying to identify the preferred policies and policy-mix (endogenous policies), what we are looking for here is the institutional setting or the mechanism, given the set of ranked policies. Hence, it is the mechanism, not the policy, that is endogenous. The specific question to ask would be: Are the selected policies and policy-mix identified in the previous chapter implementable in the sense that they can be aligned with the desired goals and accepted by all agents involved? To address this question, a particular approach based on the mechanism design theory (MDT) is used.

The bulk of this chapter is devoted to responding to two key questions of interest: Which among the preferred SC-compatible policies and policy-mix revealed in Chap. 2 are implementable in the sense that they are aligned with, and can be accepted by, MSEs and MEs; and are there an institutional setting or mechanisms that can be designed to implement them? The organization of this chapter is summarized in Fig. 3.1. We begin in the next Sect. 3.1 with the discussions on the MDT before showing its applications to our survey results and the implementability of the selected policies in Sect. 3.2.

Survey story: Farming outside Pontianak, West Kalimantan, where a famers associatian (Gapoktan) was formed to counter the middlemen so that they can get better prices for their products. Efforts to raise productivity at the production side by using digital farming did not give significant results since a more important challenge for the farmers was on the downstream side, i.e., marketing. With the help of local BI office, the association worked together with an internet-based marketing firm from which farmers received better prices. A cooperation through the network between farmers and a private institution such as this was made possible by their mutual trust and understanding about the real challenges faced by farmers, and the importance of serving local community with local products

Fig. 3.1

Organization of this chapter

3.1 Methodology: Mechanism Design Theory (MDT)

When agents have different sets of policy preference, how do we reconcile the difference such that the outcome reflecting the desired goal is achieved based upon which selected policies are implementable? If such policies exist, what kind of institutional setting or mechanism can be applied to ensure that the desired goal is achieved? This is a typical problem addressed by the MDT, where agents’ preference is reflected in his/her type ($\theta$) and decision ($d \in D$). Based on his/her strategy ($s\in S$) following that strategy, an outcome ($a\in A$) and a level of utility ($v(s,\theta )$) are obtained. For each agent i, the outcome is $a_i(d(\theta _i))$ and the strategy (message) is $s_i(d(\theta _i))$. The predetermined desired goal, often referred to as social goal, is embodied in the so-called “social choice function” (SCF), $f(\theta ) \subset A$, where $f(\theta )$ consists of optimal (best) outcome in state $\theta$. Hence, $f(\theta )$ maps a type profile $\theta$ to an outcome. Having agents sent the messages, following the MDT the equilibrium in the game can be designed to implement the social choice function $f(\theta )$.

There is the original mechanism whereby N number of agents ($N = 1,\cdots , n$) send messages ($s_1, \cdots , s_n$) that will result in optimal outcomes aligned with the SCF. The problem is that agents may not tell the truth about their type, that is, the sent messages may be untruthful, $(s'_1, \cdots , s'_n) \ne (s_1, \cdots , s_n)$. The main task is therefore finding out if there is a new mechanism through which those untruthful messages, when embedded into that new mechanism, can be “adjusted” to become truthful such that an optimal outcome aligned with the SCF can be achieved. When such a mechanism is found, there are no longer incentives for agents to lie. Sending untruthful messages will give them a lower utility. Another way to put it, whatever messages agents sent, they will be “adjusted” by the new mechanism to become truthful messages.

The question is: are there such mechanisms? If so, how do we find them? This is essentially the problem that MDT addresses, which is in contrast with a standard game theory where the rule of the game or the mechanism is already given, and the outcome of the game is what the players are looking for. In MDT, the process is the reverse: the outcome is given and the rule of the game or the mechanism is what the designers are looking for (the mechanism is endogenous).

Finding a class of mechanisms or institutions whose equilibrium outcomes implement a given set of normative goals or welfare criteria is a tedious task. It was Vickrey (1961) who first showed that if preferences are restricted to the case of quasi-linear utility functions, then the mechanism dominant strategy is dominant-strategy implementable. Advanced further by other researchers, the whole concept eventually led to the development of implementation theory, one of the central tenets of MDT that has profound implications on policy creation. According to the theory, if a mechanism has the property that each agent’s dominant strategy is to honestly report the truth, then a social choice rule (SCR) is dominant strategy incentive compatible (also termed strategy-proof).¹

In our context, agents’ strategies and outcomes are the MSMEs’ ranking of preferences toward different SC-compatible policies which depends on their state or type $\theta$ (can be unknown). While in general $\theta$ can take different forms, such as production technology, available resources, and agents’ payoffs from outcomes $a \in A$, in our case we assigned two forms of $\theta$: a time-related attribute, more specifically pre-COVID versus late COVID, and agents’ state of mind attribute, more specifically quick/fast think versus comprehensive/slow think. Given a rather complex nature of the interconnections among goals, challenges, and alternative policies or policy-mix, there is a high likelihood that the preference revealed by MSMEs varies due to their different perception toward these factors. The ranking also may not reflect the true type of MSME. If we are to find the implementable policies out of several alternatives, the information on agent’s actual type is critical not only because different states or profiles may cause different rankings of policies but also because it is necessitated by the so-called monotonicity test (see Appendix D for the explanation and example).²

Consider two groups or agents, say, MSE and ME. According to the test, if under a certain state or profile an alternative $x \in X$ is selected by one group and x does not fall in rank in any groups’ preference ordering in different state or profile, then x must still be selected. The question is, if the policy choices reflect MSE or ME preferences, what different states of mind or profiles ($\theta$) are to be used in the monotonicity test? To address this issue, we used another scenario of preference ranking. The alternative scenario was derived from the results of a sensitivity analysis applied to both the AHP and the ANP ranking discussed in Chap. 2 and this chapter.

The first scenario is a profile in which the ranking of preferences revealed by MSMEs reflects what Kahneman (2011) referred to as the “slow think” or System 2. In this scenario, agents proceeded through a sequence of steps of retrieving the memory and using their cognitive program through a deliberate, effortful, and orderly process. This is where the goals and the challenges listed in the hierarchy and the network come into play (more discussions on this later). The result of our survey is the preference ranking under this system (System 2). The second scenario is more of “fast think” or System 1, where the preferences are revealed automatically and effortlessly without considering the objectives and challenges described in the hierarchy or network; instead, they are based on impressions and feelings.³ We obtained the preference ranking of this type by running a series of sensitivity analyses on our hierarchy and network results without including the goals and challenges (they were assumed to have virtually no effect on the policy alternatives). The results essentially reflect the MSMEs’ preference ranking without taking into account the complex interrelations among components within and between the stated goals and the challenges. It is equivalent to a ranking when MSMEs evoke quick first-time reaction to the question. Hence, we have a slower System 2 that represents one state/profile (say, $\theta$) based on the construct thoughts in an orderly sequence of goals, challenges, and policy alternatives, and a faster System 1 that reflects the automatic response of MSME representing another state/profile, $\theta '$.⁴ Designed to achieve improved productivity as the desired goal, the alternatives or choices take the form of either individual policies or joint policies that are SC-compatible.

Returning to the monotonicity test, a simple illustration is as follows. Consider $f^{scr}$ as the social choice rule (SCR) of two types of players ($n = 2$). Points are assigned to each of k alternatives according to the preference of each type of player. Suppose there are two states/profiles $\boldsymbol{\theta } = (\theta , \theta ')$ and four policy alternatives ($k = 4$) in each state/profile; hence, $\textbf{X} = \{x_1, x_2, x_3, x_4\}$. The criteria for choosing the policy alternative(s) is the largest sum of points. Referring to the example below, in state/profile $\theta$ the alternative $x_2$ has the most points (6, i.e., 3 plus 3, following the Borda count), so it is optimal and chosen by $f^{scr}$. In state/profile $\theta '$, however, the optimal policy alternative is $x_1$ (it has the highest points = 6, i.e., 4 plus 2).⁵ Note that since $x_2$ falls in rankings going from $\theta$ to $\theta '$, the monotonicity condition does not require it to remain optimal. Thus, in this case, monotonicity is not violated, and the chosen policy is implementable (see again Appendix D).⁶ In the language of implementation theory, this is the case where the social planners’ SCR can prescribe $x_2$ in $\theta$ and $x_1$ in $\theta '$. This approach is called the direct mechanism (Table 3.1).

Table 3.1 Example of monotonicity non-violating case

An example of violated monotonicity condition using a direct mechanism is as follows. Suppose $x_1$ is optimum in state/profile $\theta$ (total points equal 6), but is not optimum in $\theta '$ ($x_2$ is, with total points 7). Yet, $x_1$ does not fall against any other alternative, which, according to the monotonicity condition, should remain optimal. Hence in this case monotonicity is violated, and the chosen policy is not implementable. There is no mechanism that implements the SCR (Table 3.2).

Table 3.2 Example of monotonicity-violating case

The remaining question is, how one can find a mechanism that implements the policy. This is essentially a question about whether or not it is possible to find a mechanism indirectly without knowing the agent’s true state/profile. That is, the designers have only the messages, i.e., the ranking under each state/profile, but not the state/profile itself.

In the direct mechanism example above, the designers ask the players directly about his/her individual type. If, on the other hand, a mechanism can be found without necessarily knowing the agents’ type, that is, agents are asked to send only the ranking (messages and outcomes), not their state/profile, it is called the indirect mechanism. Applying to the above $k=4$ and $n= 2$ case, such a mechanism is represented by the following game, where the moves of MSE are up (U) and down (D), and ME’s are left (L) and right (R) (Table 3.3).

Table 3.3 First game example

Suppose the actual state/profile (not the reported one) is $\theta$. For ME, no matter what MSE chooses, the best strategy is taking left (L) because for them $x_2 \succ x_3$ if MSE picks U, and $x_4 \succ x_1$ if MSE picks D. Hence, L is the dominant strategy for ME in $\theta$. Since $x_2 \succ x_4$, the best strategy for MSE is U. That is, $x_2$ is a Nash equilibrium. If, on the other hand, the true state/profile is $\theta '$, by using a similar procedure we can establish that ME plays R and MSE plays D, and the Nash equilibrium is $x_1$. Therefore, this mechanism works in both $\theta$ and $\theta '$, suggesting that social planners do not need to know the true state/profile, and yet the resulting outcome is optimal no matter what that state/profile is. Since the Nash equilibrium outcomes of the mechanism coincide with the optimal outcomes in each state, we can establish that the mechanism implements the social planners’ SCR in Nash equilibrium.

Identifying this indirect mechanism is very important because in the direct mechanism shown earlier, even if social planner can ask MSE and ME directly, they may be untruthful. In this example, although the true state/profile is $\theta$, MSE will tend to say that it is not. They are likely to say it is state/profile $\theta '$ because they always prefer $x_1$ to $x_2$. On the other hand, ME always declares state/profile $\theta$ as the true one, even if it is actually $\theta '$ because they always prefer $x_2$ to $x_1$. Such a risk is eliminated when an indirect mechanism can be implemented.

At this juncture, it is important to emphasize another critical principle in MDT known as the revelation principle, which is central to the analysis of implementability. It also addresses the distinction and relation between direct and indirect mechanisms.⁷ The principle essentially says that, if an SCF can be implemented by an indirect mechanism, then it can also be implemented by a truth-telling direct revelation mechanism. Thus, when considering implementation in dominant strategies, it is enough to look only at the SCF for which truthful is a dominant strategy. We can therefore consider only truthful mechanisms and be assured that such a mechanism exists, because any SCF that can be implemented by any mechanism can be implemented by a truthful direct mechanism.

The scheme in Fig. 3.2 captures the discussions above. Given agents’ type $\theta _i$ and the strategies they choose ($s_i(\theta _i)$), when we apply the original mechanism the outcome would be ($a_1, \cdots , a_n$) and the agents’ utility is $u_i(a,\theta )$. If agents are being untruthful by using $s'_1(\theta _1), \cdots , s'_n(\theta _n)$, embedding these untruthful strategies into the original mechanism will give us a new mechanism. This new mechanism adjusts the untruthful information to become truthful such that there is no point for agents to be untruthful. That is, the mechanism will adjust the information. Thus, the revelation principle allows one to solve for an equilibrium by assuming all agents truthfully report their type subject to an incentive compatibility constraint, eliminating the need for agents to consider either strategic behavior or lying. Note, however, that the equilibria generated by indirect mechanism are not always the same as those generated by direct mechanism. They are the same if agents are truthful. To the extent the outcome generated by indirect mechanism can be bad equilibria, indirect mechanism cannot (inherently) be better than direct mechanism.

The revelation principle, which implicitly suggests that truthfulness is not a restrictive assumption, is extremely useful since the designer or the social planner does not have to consider all possible strategies (games) that agents may take and choose one that best influences other agents’ strategies to align with the SCF. Instead, they can simply consider games in which agents truthfully report their private information (direct mechanism).

Fig. 3.2

Summarized system of MDT

Fig. 3.3

Summarized relationship of monotonicity test and mechanisms

Figure 3.3 summarizes the relation between the monotonicity test for implementability and the two types of mechanism: direct mechanism and indirect mechanism. The results of monotonicity test are obtained by applying an incentive-compatible direct mechanism in which being truthful (T) is the dominant strategy. In some cases, an indirect mechanism can be established, where sent messages do not have to be truthful. Any implementable strategy in such a mechanism is also implementable in direct mechanism (following the revelation principle), although the resulting equilibria from the indirect mechanism can be either the same or different from those generated by the direct mechanism (Equilibria* $\ne$ Equilibria). They are the same if agents are truthful (T), or can be different if some of them are not truthful (U). If the latter holds, the results of indirect mechanism can be bad equilibria, implying that indirect mechanism cannot be better than direct mechanism.

Since it has been proven that any equilibrium outcome of an arbitrary mechanism can be replicated by an incentive-compatible direct mechanism, the optimal mechanism can always be found within a sub-class consisting of direct mechanisms (Myerson, 1986). The discussions on implementability and the monotonicity condition in the next section use largely the direct mechanism, although in some cases we also show the indirect mechanism to identify which SC-compatiblepolicies or policy-mix that are in equilibrium and implementable.

3.2 Implementability

From the discussions in Chap. 2, we found that our MSME respondents placed “Interaction-network” and “Supporting infrastructure” as the most important SC-compatible policy measures for productivity improvement. This is consistent with how they ranked the policy-mix, in which a combination of policy to create a network and improve infrastructure was selected as the one having the highest priority. The question is, are these policies and policy-mix implementable in the sense that they are incentive-compatible and aligned with the social choice function (SCF)?

As indicated earlier, we need to conduct monotonicity tests to check the implementability of policy choices based on the hierarchies in Figs. 2.6 and 2.7, as well as on the networks in Fig. 2.5. Before conducting such tests, we first need to explore the truthfulness of the state/profile under which MSMEs made such rankings. One relevant issue to explore is whether the ranking revealed by MSMEs during the survey was made after they carefully analyzed the objectives and the challenges (‘think slow’, or System 2), or was it made based on their quick and automatic response to the questions with minimum efforts (‘think fast’, or System 1). The former state/profile is denoted by $\theta$, and the latter by $\theta '$.

We are also interested in another scenario involving panel data. More specifically, we wish to find out whether or not MSME perceptions changed after the devastating COVID pandemic. Have they become more pessimistic and anxious, or more positive, holistic, ethical, and environmentally conscious in contemplating their role or life in general? The state/profile when the COVID hit is denoted by $\theta '$ (2020 survey) and by $\theta$ (2022 survey) when the pandemic has abated. For that purpose, we used the preference ranking revealed by the same respondents who participated in both surveys and made the questionnaires in the two surveys to be comparable.

There were five alternative choices in the 2020 survey, comprising of three policies and two social capital capable of influencing cooperation for collective action. The three policies consist of: one, to promote and support linkages or interaction between MSMEs and the relevant stakeholders; two, to provide financial and technical support and launch promotion; and three, to use digital and green technologies.

Survey story: Aiming to empower local women, improve their quality of life and revive the local weaving tradition, an MSME in Kupang, East Nusa Tenggara, employed local weavers like this to utilize their skills by using local materials. By building a network of many women weavers, the MSME was able to help transforming the weaving tradition from something with only cultural values to a source of income for local women. During our survey, we found several similar cases in other areas throughout East Nusa Tenggara

For the social capital part, the two components that support cooperation are participation and coordination. Replicated from Fig. 4.4 of Azis (2022), Fig. 3.4 shows the network capturing those five alternative choices, their interactions including feedback effects, and the contents (elements) of each of those choices. We matched five alternative choices with the six policy measures in the 2022 survey shown in Fig. 2.5 earlier. For the ranking of preferences, in the first case we used the results obtained from the 2020 survey since they were exactly the same five policy choices, and for the 2022 case we combined “Interaction-network” and “Regulation & legal matters” by taking their average weight before normalizing the priority ranking.⁸

Fig. 3.4

Interplay of policies and social capital: A network of feedback and interrelations

Table 3.4 Monotonicity test for direct mechanism: Did COVID alter MSMEs’ perceptions?

The results of the concordance between the two sets of policy choices and the corresponding ranking of preferences are listed in Table 3.4. The ranking for ME and MSE according to the 2022 survey is shown in the first two columns (under state/profile $\theta$), and the ranking from the 2020 survey is in the next two columns (under state/profile $\theta '$). Notice that the two rankings were not the same. While the top-ranked policy was always “Linkage/Interaction,” the second- and the third-ranked choices were different, depending on whether or not the respondents considered the complex relations between objectives, criteria, and policies when revealing their preference. If they did ($\theta$), ME and MSE picked different choices such that the following pairs were for the second and the third rank, respectively: “Coordination” with “Technology” and “Structural” with “Coordination.” On the other hand, if all the complex relations were ignored ($\theta '$), implying that the respondents gave their direct ranking of policies instantly, the selected pairs were exactly the same, i.e., both ME and MSE picked “Technology” and “coordination” for the second and the third ranks, respectively.

What about the optimal choice of both players? If the true state/profile is $\theta$, both MSE and ME found the “Linkages/Interaction” policy to be optimal, where the weight sum equals to 0.537 (0.260 plus 0.277). The same policy is found optimal if the true state/profile is $\theta '$ where the weight sum is 0.536. Thus, social planners know what policy to take regardless of the true state/profile. We can therefore surmise that the shock due to the COVID pandemic did not alter MSME perceptions in assigning the highest priority to the preference to have a network. Most MSMEs were of the opinion that being involved in a network would enable them to have more interactions (quantity of network) and greater linkages (quality of network) with the stakeholders.

Next is to check the implementability of SC-compatible policies based on the network in Fig. 2.5. The state/profile we used refers to the circumstances under which MSMEs revealed their preferences. The fact that the ranking of policies obtained from the survey was based on the complex interactions involving feedback effects between objectives, challenges, and policies, it reflects System 2 where they essentially applied the cognitive program through a deliberate and orderly process. This is denoted by $\theta$. But there is also a scenario where the state was much simpler: i.e., they revealed the preferences by using System 1 based on their impressions and feelings, without considering how those preferences were related to any of the objectives, challenges and anything else. Such a state/profile is denoted by $\theta '$. It is not surprising that the resulting ranking was different, in that the equilibrium outcome under $\theta$ was “Interaction-network,” while that under $\theta '$ was “Supporting infrastructure”; see Table 3.5. This finding is unlike the earlier results when we tested the possibility of a change in MSMEs’ perceptions due to the COVID pandemic, where “Linkage” being the optimal outcome in both $\theta$ and $\theta '$.

Table 3.5 Monotonicity test for direct mechanism: Preferred policies of ME and MSE

Does this mean “Interaction-network” is no longer implementable? In the above case, social planners know what policy to take given a particular state/profile. What about the case if they do not have such information? This is where monotonicity test can be helpful. In moving from state $\theta$ to $\theta '$, the ranking of “Interaction-network” falls from the first to the fourth. According to the monotonicity condition, a different equilibrium from “Interaction-network” can be optimal. In this case, the new optimal SC-compatible policy is “Supporting infrastructure.” Since the monotonicity condition is not violated, there is a mechanism that implements SCR. By considering only four SC-compatible policies (the remaining two received the lowest ranking), Table 3.6 shows such a mechanism. Notice that no matter what the true state/profile is, if $\theta$ is the true state the Nash equilibrium outcome is always “Interaction-network” as MSE always prefers strategy U and ME prefers strategy L. Similarly, when the true state/profile is $\theta '$, the equilibrium outcome is always “Supporting infrastructure.”

Thus, even without knowing the true state/profile, the predicted outcome or Nash equilibrium based on the preferred strategies of MSE and ME is the same as the desired (social) outcome. We can therefore establish that the mechanism implements the social choice rule in Nash equilibrium because the equilibrium outcomes of the mechanism coincide with the optimal outcomes in each state/profile.

Table 3.6 Indirect mechanism to align MSE and ME preferences with SCF under pre-COVID and abated COVID state/profile

The next to test is the set of SC-compatible policy-mix. By using the same specifications of the state/profile as before (System 2 for $\theta$ and System 1 for $\theta '$), the following Table 3.7 shows the preference ranking.

Table 3.7 Monotonicity test for direct mechanism: Preferred policy-mix of ME and MSE

Recall that in any policy-mix cases, the results identify the preference ranking of pairs of SC-compatible. Based on the weight of each pair, the combination of “Interaction-network” and “Linkage” is the most optimal under $\theta$, and the same pair is also optimal under the state/profile $\theta '$. Since the ranking of that pair is not altered from $\theta$ to $\theta '$, this optimal policy-mix is implementable. Thus, the SCR prescribes “Interaction-network & Linkage” in each state/profile.

Similar to the preceding case, even without knowing the true state/profile the predicted outcome or the Nash equilibrium based on the preferred strategies of MSE and ME is the same as the desired (social) outcome. The indirect mechanism that supports such a conclusion is shown in Table 3.8. If the true state/profile is $\theta$, ME would prefer strategy L expecting to get “Interaction-network & linkage” because that policy-mix is ranked at the top. The choice made by MSE is to take the U strategy because as shown in Table 3.7, the “Interaction-network & Linkage” is preferred to “Interaction-network & Promotion.” On the other hand, if the true state/profile is $\theta '$, ME would take strategy R to get “Supporting infrastructure & Linkage,” which is ranked the highest, and MSE takes strategy D to get the same policy-mix. Hence, regardless of the true state/profile, the mechanism always gives a Nash equilibrium outcome that coincides with the outcome obtained from using the direct mechanism.

Table 3.8 Indirect mechanism to align MSE and ME preferences under state/profile System 1 and System 2

Reviewing the precise wording (in Bahasa) used during the survey, in the network-based questionnaire most of the statements were very closely related to having a network with the stakeholders, such as increasing the number of business links (“relasi bisnis”) and getting the benefits from cooperating with other businesses, suppliers, supporting industries, communities, and government apparatus (“bantuan kerjasama dengan bisnis lain, supplier, industri pendukung, masyarakat sekitar, dan perangkat pemerintah”). In the hierarchy-based questionnaire for policy-mix, the corresponding wordings were similar and even broader: increase business relation (“Memperbanyak relasi bisnis”), get access to participate in the supply chain involving other firms including large and other businesses (“Bantuan pengingkatan akses untuk bisa berpartisipasi dalam rantai pasok dengan bisnis lainnya maupun dengan perusahaan besar dan bisnis pendukung”), and increase the competitiveness (“untuk meningkatkan daya saing”), including guidance and consultation in marketing, administration, technology, including green and digital technology, and financial matters (“bimbingan dan konsultasi terkait strategi pemasaran dan administrasi ataupun hal lainnya seperti adopsi teknologi baru termasuk teknologi hijau dan digitalisasi, dan masalah keuangan”).

Looking at all the results from monotonicity tests, it is noteworthy that in every single case of our survey, both the hierarchy and the network-based, the implementable SC-compatible policy is to support greater interactions through a network. For the policy-mix, the highest-ranked choice is further bolstered by the preference for linkages with other stakeholders. Even under a hypothetical case of respondents answering the questionnaire based simply on their impressions without considering the objectives and the challenges (System 1), where supporting infrastructure is the preferred choice, they tend to associate that choice with strengthening the linkages. Clearly, having a network is viewed by MSMEs as a vital step for improving their productivity because through a network they gain more interactions and better linkages among themselves and with other stakeholders.