**3. Results**

Following the grounded theory process outlined above, a system was formed involving 279 coding reference points, 19 basic categories, and 4 core categories as the explanatory model's basis. In this model, the nodes at all levels from the bottom to the top have subordinate relationships, and were formed through a bottom-up induction process.

#### *3.1. Open Coding—Primary Related Factors of RM*

The concept genera, or set of primary factors related to RM, was derived from the 124 selected papers, and named from the original texts based on grounded theory's "localization" principle. 19 three-level nodes were obtained by combining and coding the papers' contents. From this, the following direct factors were identified:


#### *3.2. Associative Coding—A Structural Model for RM Research*

Associative coding was conducted to further discover and establish the relationships between conceptual genera and their organic correlations. This provided an initial set of 11 s-level nodes, some of which were inferred from the induction and integration of qualitative analysis based on third-level nodes. They represent intermediate elements in the RM analysis. Further qualitative analysis of the 11 secondary nodes revealed four primary nodes located at the top of the subordinate relationship, and embodied macroscopic elements for RM study. Figure 4 shows the structural model.

**Figure 4.** Structural model for RM analysis. Second-level nodes are outside. First-level nodes are inside.

The center of Figure 4 contains the model theme, namely RM. The specific hierarchical structures of the RM-related research elements are represented by two-layer circles. The wedges within each circle reflect the dimensional categories (nodes). As Figure 4 illustrates, current RM studies consist of 4 first-level nodes and 11 s-level nodes. Third-level nodes are not represented due to their large number. The size of each ring sector is determined by the number of coded reference points included. This represents the "volume" (importance) of papers supporting each node and indirectly reflects their influence.

As Figure 4 shows, the *Focused issues* node accounts for most reference points among all first-level nodes. In this category, RM is a popular topic for different effect analyses. Indeed, the analysis of misallocation based on the construction market is also supported by the majority of reference points among all second-level nodes. This indicates that this market environment is still a key factor for improving current RM situations. In particular, the market economy of China's construction industry is currently in the process of relatively long-term adjustment of the industrial structure and needs to be explored further in subsequent studies. In contrast, the *research deficiencies* node contains the least number of reference points and also has the least influence.

#### *3.3. Selective Coding—The RM Research Core*

In the final stage, focused issues and misallocation degree measurement and characterization were determined as core categories. These two core genera include 7 subgenera, involving 235 coding reference points and account for 60.2% and 24.0% of the total analyzed, respectively. These two core genera occupy the main position among all genera and represent important links in RM studies. Figure 5 summarizes the final data structure of our selective coding research.

**Figure 5.** Selective coding data structure.

*Focused issues* include three specific indicators of upgrading of construction industry: structure, organization management, and effect analysis. Of these, there are a total of 83 effect analysis nodes with the largest number of reference points (49.4%). This again shows that research into the effect analysis system's direction the plays a key role in guiding the effective allocation of resources. Improving the effect analysis system will further improve (optimize) the allocation of corporate resources. Moreover, the *measurement method* contains three specific index points: the simple ratio method, variable substitution method, and growth rate decomposition method. Of these, there are a total of 52 variable substitution method nodes with the largest number of reference points (77.6%) of the effect analysis node. This shows that the measurement method using the variable substitution method for robustness tests reflects the RM level being measured in many practical settings. Conversely, *sources and concepts* and *research deficiencies* are relatively minor.

Moving forward, we focus on the core and supporting *genera*'s high-frequency nodes of *sources and concepts*, *focused issues*, and *misallocation degree measurement and characterization* for the model's explanation. *Research deficiencies* represents a very low-frequency node and is not developed further.

#### (1) Sources and concepts

As Table 1 shows, *source and concepts* can be considered the entry point in current research.


**Table 1.** Key literature reviews of sources and concepts.

The sources and concepts node includes three secondary nodes: theoretical origin, research beginning, and research reasons, with a total of 23 reference points. Of these, there are 13 theoretical origin nodes, which contain most reference points and occupy a relatively important position. At the beginning of the research, there are 5 reference points and the reason for the research occupies 5 reference points. The number of points included is small and its impact is small too. The number of reference points for each secondary node is equally small, as is their gap, indicating that the source of RM and the concept's dimension have received little attention (apart from the theoretical origin).

In general, the allocation of resources at both macro and micro levels will be continuously optimized and updated due to the continuous development and change of external market environments. Therefore, its sources and concepts are often ignored or not widely mentioned. Table 2 shows the number of nodes at all levels and coding reference points included in sources and concepts node.


**Table 2.** Number of nodes at all levels and coding reference points included in the sources and concepts node.
