2.1. Drivers of related diversification

Related diversification is basically explained by efficiency motivations. Synergies between production processes may increase efficiency and motivate firms to diversify into related businesses [40–42]. Several studies suggest that related diversification is superior to unrelated diversification when the primary goal is to increase the efficiency, because synergies increase firm performance [43–46].

This argumentation highlights that multi-product firms make more efficient use of resources [47]. The underlying idea appears already in Penrose’s [1] explanation for corporate growth: if firms have indivisible resources, which cannot be fully exploited in their original core business (excess capacity arises) and these resources can be utilized in other industries, it encourages companies to grow through diversification. The theoretical foundation of the underlying cost-efficiency arguments is the concept of economies of scope [22, 48–50]. According to Panzar and Willig [50], the source of economies of scope is the sharing or joint utilization of inputs. It arises if a given input is imperfectly divisible and its utilization in producing a single product would leave excess capacity. Further, certain resources, such as information, skills or knowledge do not deteriorate; therefore, they are also available freely for other production processes. Indivisible (or not fully divisible) resources incur fixed costs (or quasi-fixed costs) [22, 50, 51]. If these fixed costs are divided across various products in case of multi-product firms, then they result in lower unit costs.

Teece [2] emphasizes that the resource-based explanation of diversification always assumes the presence of transaction costs. If the product and capital markets worked perfectly, these efficiency aspects could not justify the appearance of multi-product firms. If there were no transaction costs, excess capacity of indivisible resources could be efficiently sold or leased in the market, thus the cost advantages behind the economies of scale and scope could be exploited in the same way for single-product firms.

Another economic mechanism explaining the efficiency of joint production of products within firms is complementarity. Weiss [52] emphasizes that the common notion of equating relatedness with similarity is incomplete because it does not take resource complementarity into account when the joint use of distinct resources produces a higher total return than the sum of returns of independent utilization of each resource. Consequently, firms primarily search for complementary resources rather than similar resources during acquisitions, because combination and integration of the acquirer’s resources with target firm’s resources might increase the efficiency of resource utilization [53, 54].

2.2. Drivers of unrelated diversification

Explanations of unrelated diversification can be classified into three major groups: risk reduction motives, motivations to increase market power, and individual managerial motives.

Several studies [55–58] argue that in case capital markets are not efficient, there may be financial motives for unrelated firm expansion, that is, the product diversification can be an element of risk reducing strategies. Firms can allocate their activities across industries with opposite market risks to reduce the exposure of their core business. Amit and Livnat [59] revealed that benefits of unrelated diversification include more stable cash flows, lower operating risks and increase the levels of leverage. Galbraith et al. [27] showed that in R&D-intensive industries unrelated diversification is frequently applied as a hedging strategy against unforeseen technological shocks. Like the economies of scope theory, this explanation also requires market failures and transaction costs in capital market, otherwise these risks can be reduced by allocating investments among various financial instruments. Indeed, Dundas and Richardson [29] revealed that market failures in product and technological markets motivate the firms to diversify into related industries, whereas imperfections in capital markets raise the probability of unrelated diversification. Robson et al. [60] observe a similar risk-sharing motive of diversification in small firms. They argue that diversification is often used by small firms as a risk-spreading survival strategy rather than a growth strategy. The authors find a negative relationship between diversification and growth in the case of small firms. They conclude that small entrepreneurs do not have the necessary resources and skills to manage diversified activities and therefore diversification can be considered a risk-sharing strategy in small firms, these entrepreneurs tend to diversify to compensate a decline in their original core business.

Other explanation of unrelated diversification is firms’ strategy towards increasing market power. The literature of conglomerates emphasizes the anti-competitive effects of the formation of large multi-industry companies operating in unrelated industries [61–64]. According to these explanations, operating in unrelated industries allows companies to engage in anti-competitive practices that include cross-subsidization, when the profit earned in one market allows maintaining predatory pricing in another market [57], and mutual forbearance, when big conglomerates compete in multiple markets and recognizing this interdependence, they mutually respect each other’s dominant markets, competing less vigorously [64].

Finally, explanations of other individual managerial drivers are based on the agency problem arising from the separation of ownership and management. Shleifer and Vishny [65] present that managers use strong corporate expansion as a tool to increase the firm’s demand for their personal skills. Amihud and Lev [66] argue that managers also use corporate expansion (unrelated conglomerate mergers) to reduce their employment risk. While corporate owners can reduce their risks by creating sufficient capital market portfolios, corporate executives can only reduce their employment risk by expanding the company’s activities. Hence managers use diversified expansion as a means of reducing total firm risk and at the same time their employment risk and smoothing the earnings stream, which could lead to excessive diversification. Accordingly, they find that manager-controlled firms are more diversified than owner-controlled firms and that the conglomerate type (i.e. unrelated) expansion is more frequent in case of manager-controlled firms than owner-controlled firms. Denis et al. [32] show that managerial ownership is negatively related to diversification, and that market disciplinary forces, such as managerial turnover or financial distress increase corporate focus. Gomez-Mejia et al. [67] finds that family firms diversify less than non-family firms.

2.3. Firms’ strategies when facing negative shocks

Firms can choose different strategies when facing difficult economic conditions due to environmental shocks, economic recessions, and declining industries. Kitching et al. [30] distinguish an investment-based expanding strategy, as against the cost reducing adaptation to crisis. As they show, some firms perceive crisis periods as opportunities to invest, innovate and expand into new markets. However, most firms are not able to implement such investment-based expanding strategy, because it requires financial and managerial resources. Firms hit harder have limited resources and focus more on short‐term survival, choosing cost-cutting strategies more likely [30] and in fact, the most common behavioral response of firms to falling demand is cost cutting [68]. These retrenchment strategies can consist of a wide range of cost reducing actions: divestment of assets, establishment closure, labor cost reduction (downsizing or working hours reductions), cutting expenditures of non-core activities such as R&D, marketing, and employee training [30]. This strategy is the typical for small firms [69], and it also seems to improve their performance [70].

Despite cost reduction is the most frequent strategy of firms when facing negative demand shocks (e.g. in times of economic downturns), some authors debate its’ effectiveness. They argue that those firms that cut costs in order to survive in the short‐run must face long-run opportunity costs of capacity reduction, because they might be unable to adapt adequately during economic recovery [71]. In addition, when demand is lower, opportunity costs of investing in productivity-enhancing innovations may be lower [72], and in crisis times payoffs from investments like R&D and marketing may be higher, as there is lower level of such activity on the market due to budget cuts [73].

Increasing import competition can be another cause of industry-level demand shocks. Imported products are substitutes for domestically produced products, causing firms to face declining demand for their products and stronger price competition. Several studies examined the impact of increased exposure to imports from low-wage countries on labor market outcomes. It was found that increasing imports from China is negatively associated with the level of employment, labor force participation, and wages in import-competing manufacturing industries [74–77]. Bernard et al. [78] find that survival and growth rate of those US manufacturing plants decreased which has been exposed to increasing industry-level import competition from low-wage countries. Bloom et al. [79] show that increasing Chinese import competition has led to a decline in employment, profits and prices. Dobson and Waterson [38] show that intensified competition may encourage firms to specialize and thus limit their product range. There are also studies however that highlight the positive effects of Chinese imports on European firm performance. Such effects may arise across different product groups within the same firm through providing cheaper but high-quality intermediate inputs thus facilitating a more efficient production [80–82].

2.4. Research questions and hypotheses

Previous studies suggest that firms frequently choose cost reducing retrenchment strategies as a response to the falling demand and expected profit, decreasing cash flow and lack of financial resources [30, 68–70]. Accordingly, we can suppose that facing demand shocks makes the cost efficiency concerns more important for firms, which will limit the options of managers. In the context of product diversification, efficiency-based drivers – the needs for more efficient use of resources – encourage utilization of economies of scope which leads to related diversification, while in the case of non-efficiency driven corporate expansion (driven by risk reducing motives or individual managerial and market-power-based drivers), technological proximity of industries is either irrelevant or even encourages unrelated diversification.

We can thus expect that firms rather choose the exploitation of economies of scope, while other drivers of diversification (risk-sharing concerns or other managerial motives of firm expansion) are pushed into the background when they are exposed to negative demand shocks. To formulate a hypothesis around these claims, we measure the technological coherence of product portfolios by tools of technological relatedness such that a product portfolio is coherence in case it contains technologically related products. Negative demand shocks are measured by decreasing markets and increased import competition. These measurements are described in the next section and captures the potentials of economies of scope.

1. H1a: When exposed to negative demand shocks, the technological coherence of company portfolios increase.

However, opportunities of managers to choose between different strategies may also be influenced by endowments of the firms. Therefore, we expect that firms that have less tight liquidity constraints are less pressed towards exploiting the short-term efficiency benefits of the related product portfolios. Therefore:

1. H1b: Demand shocks increase technological coherence of product portfolios more intensively in those firms that face liquidity constraints.

When negative shocks hit in, firms’ strategies can include retrenchment and expansion as well. Thus, one can investigate the dynamics of product portfolios and distinguish events of dropping and adding products. Based on the previous argument, we expect that efficiency arguments will gain more importance in both types of crisis situations compared to prosperous times. We can formulate our hypotheses accordingly:

1. H2. When exposed to negative demand shocks, firms are more likely to drop products that are technologically less related to their core production than more related products.
2. H3. When exposed to negative demand shocks, firms are more likely to add more products related to their core production than unrelated products.

3.1. Data

We use the Hungarian PRODCOM database for years 2003-2012. PRODCOM data is based on reporting the production volumes and sales values for each product the firm produces to the Statistical Office on a yearly basis. This reporting covers (1) all companies that have more than 20 employees operating in industrial sectors, (2) a sample of industrial companies employing 6-19 workers, and (3) all firms having more than HUF 500 million (€ 1.65 million) sales revenue from industrial products, even if they are not classified to the industrial sector by their primary activity. This results in a sample of approximately 7,000 firms and 2,400 products (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Number of firms and products in the analysis.

https://doi.org/10.1371/journal.pone.0280461.g001

This sample in terms of number of firms seems to be relatively small compared to the number of all firms operating in Hungary, that were around 550,000 in the observed period (EUROSTAT). Approximately 90% of Hungarian firms are however very small, having less than 5 employees (STADAT, Hungarian Statistical Office [83]; therefore, they are not included in the PRODCOM database. Comparison of the PRODCOM data to all enterprises with more than 5 employees (Fig 2) reveals that it indeed covers more than 90% of the medium and large firms operating in the manufacturing and in the mining sectors, and more than 50% of the firms in the 10-50 employee category. Distribution by industry shows that 95% of the observations come from manufacturing industries. Despite firms from the mining and from the energy and water sectors also have high probability of being included, they represent a small minority (3%) of the data, due to the relatively low number of such firms in Hungary compared to manufacturing. Firms registered in other sectors (agriculture, construction, trade and services) constitute only 2% of the observations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Number of observations according to firm size and industry with comparison to all firms.

Note: Observations represent mean values for 2003-2012. Source: Authors’ calculations based on the PRODCOM database and the balance sheet data provided by the Hungarian Statistical Office.

https://doi.org/10.1371/journal.pone.0280461.g002

Although the typical firms in the data employed 11-50 workers, due to the presence of large firms, the average firm size was 102 for the firms in our data. Firms created 72.9 thousand USD value added per year per worker on average. They operated with a 3% profitability ratio (compared to their total revenue), and on average 38% of their equity was coming from short-term debts (Table 1). Firms are relatively young, operating for 11.5 years on average, because 98% of them were established after the transition in 1989.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Key characteristics of the firms.

https://doi.org/10.1371/journal.pone.0280461.t001

Data to the import competition variable were accessed from the following sources:

1. Trade value of imports from China to EU28 countries on the HS2002 6-digit product level for years 2002-2006, and on the HS2007 6-digit product level for years 2007-2011 from Comtrade.un.org.
2. EU28 Consumption was proxied as Sold production (Sales) + Import – Export, for which data was accessed from Eurostat’s Prodcom database.

3.2. The measurement of technological relatedness

To measure technological relatedness between products, we apply the revealed relatedness methodology developed by Neffke and Henning [84]. This method calculates relatedness measures between industries and assumes that they are technologically related in case firms often produce the combination of products that belong to these industries.

Relatedness between products is calculated on the industry-level, which is equivalent to the first 4 digits of Prodcom product codes that define the 4-digit NACE industry codes. Therefore, industries can be viewed as product groups. This level of aggregation makes the results comparable with previous studies using this method [20, 84, 85]. Calculating relatedness across 6-digit Prodcom codes would introduce too much noise to the measurement and the indicators would turn highly volatile when decomposing it to years. Accordingly, in the first step of our data procession, we assign the products of the firms to industries.

Next, the core product of each company is identified. We define this as the product that the biggest part of the firm’s total revenue stems from. We applied this definition on yearly basis; therefore, it is possible that a firm changes its core product over time. Formalizing the definition, let Y_ipt stand for the sales value of firm i from product p in year t, then for firm i across its products p, the firm’s core product in year t is c_it = p, where Y_ipt = max_it (Y_ipt).

Third, we define the industry of the firms: each company is assigned to a 4 digit industry code, which corresponded to its core product in the PRODCOM classification.

This classification enables us to calculate technological relatedness of industries A and B as the frequency that firms produce two or more products such that at least one belongs to industry A and at least another one belongs to industry B. A directed link from industry A to industry B is created, if a company’s core product belongs to sector A, but it also produces products, which belong to sector B according to the PRODCOM classification. By summing the links over the industries, we got the co-occurrence network of industries, where each node corresponds to a 4-digit industry, and each edge weight corresponds to the number of companies active in production in both industries.

As we discussed above, diversification into products can be motivated not only by economies of scope and technological proximity. Other factors, such as high expected profits or industry size are important as well so that firms might decide to expand their activities to attractive industries. To control for other factors that influence the number of co-occurrence links we follow Neffke and Henning [84] and estimate the expected number of co-occurrences for each pair of industries using industry specific characteristics as predictors (e.g. profitability, sales volume, competition expectations, wage level etc.). Then the difference between the predicted and observed co-occurrences will reveal the relatedness between each industry pair concerned. As most of the potential pairs of industries do not occur along with each other in the same firm, i.e. there is an extensive number of zeros, we follow Neffke and Henning [84] who suggest a zero-inflated negative binomial regression to estimate the level of co-occurrence. The model is formulated as follows: (1) where L_ij is the number of actual co-occurrence links from industry i to j and v_i and w_j are the specific characteristics (number of active firms, average profit level, income level, firm size in employees, and value added) of industry i and j.

From the predicted values of the regression, we could get our revealed relatedness indicator: (2) where ^ indicates fitted value and k is a normalizing constant.

The matrix of RR indices can be visualized as technological proximity network of industries that we use to explain firms’ diversification behavior and the dynamics of the product portfolio during crisis (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The technological proximity network of manufacturing sectors.

Nodes depict manufacturing sectors according to NACE (Rev. 2) 4-digit level codes, edges represent the average revealed relatedness value for the period 2003-2012. Color codes have been set according to NACE (Rev. 2) 2-digit industries. Position of the nodes is determined by the Force Atlas 2 algorithm of Gephi [87].

https://doi.org/10.1371/journal.pone.0280461.g003

One might raise the concern that such a revealed technological relatedness measure might be distorted by the presence of: 1) conglomerate firms that choose to differentiate their production lines in structurally unrelated markets; 2) vertically integrated firms that choose to produce the inputs necessary for their production. Presence of conglomerated and vertically-integrated firms would not be a problem if we had site-level data, as firms produce technologically related products at the same site, which can benefit from the resulting cost efficiency in production. Unfortunately, our database only contains firm-level data. However, two things reduce the problem raised. First, to avoid this multi-plant problem, our analysis focuses on the manufacturing sector. Békés and Harasztosi [86] show that in the Hungarian manufacturing industries over 90% of firms have one site only, which suggests that the bias is not significant. Moreover, as shown above, the difference between the predicted and observed co-occurrences reveals the relatedness between each industry pair. The relatedness indicator between two industries would be significantly distorted if both industries were composed of very similar conglomerates or vertically integrated companies expanding in the same unrelated directions. Thus, the problem is considered legitimate, but the arguments above suggest that the bias does not really give cause for concern.

The technological proximity network of Hungary (Fig 3) reveals that those 4-digit industries tend to cluster that belong to the same 2-digit NACE industries (we made available the technological proximity data at: http://www.mtakti.hu/relatedness/). For example, majority of subsectors in the food industry cluster together in the upper part of the network. Additionally, one might also find logic behind the relative position of industries in the network in two terms. First, there is a relatively dense core of the network including machinery, basic metals etc; while there are loosely connected peripheries in the network including food, apparel, textiles, wood etc. industries. Second, those industries that one assumes to be similar by intuition (e.g. wood and furniture) are indeed neighbors.

3.3. Measuring diversification

The technological coherence of firms’ production portfolios is the main object of the first hypothesis. For the measurement, we identify the core product of the firm as the product that the biggest part of the firm’s total revenue stems from. The coherence of firms’ product portfolios is defined as the arithmetic mean of relatedness between the firm’s core product and all additional products: (3) where c is the core product of firm i (for simplicity from here on we omit the it indices from the notification of the core product and only refer to it a c instead of c_it.), m is the number of products the firm produces, p≠c represents its non-core products, Y_ipt is the sales value of firm i from each product p, and RR_cp,t−1 is the relatedness index between the industries of product p and c in the previous year.

We apply the lag of relatedness that is necessary to avoid potential endogeneity. Since we calculate relatedness from average product portfolio of firms, the measure without time-lag would be problematic to use to explain the decision of firms on product portfolio. The time lag allows us to examine how the average relatedness index of the product portfolio measured in the previous year changes in the current year, for example, in response to a demand shock.

3.4. Measurement of negative demand shocks

Our analysis focuses on the relationship between product-level demand-shocks and production portfolios, and thus we identified our measures on the product level. At the first instance, we define product-level Market shrinkage based on production volumes. Secondly, we define product-level Import shock based on increasing share of Chinese imports.

For simplicity, we measure the market shrinkage using the relative change of its production value from the previous year over all the producing firms. Thus, our product-level measure is: (4) where Y_pt is the sales value of all firms producing product p in year t. Note that more severe market shrinkage is indicated by positive values, while negative values indicate expansions of the market or prosperity. To measure a firm’s exposure to decreasing markets, we calculate firm-level (weighted) averages of the product-level crisis over the m products, a firm produces: (5)

Note that we measure market shrinkage (or prosperity in case of negative values) based on production value of the Hungarian market, which also includes the observed firm. This creates mechanical correlation between our dependent and independent variables. To avoid potential endogeneity, we will use the lagged version of our crisis measure to predict diversification in the following year. However, if the firm’s decisions are correlated over time, or unobserved factors’ influence them, this still can be problematic. We have no such concerns about our other measure the exposure to import competition from China.

To measure import shocks, we follow the approach of Bloom, Draca and van Reenen [79] to proxy the shocks by the increasing export competition from China. We utilize that Hungary is a relatively open economy; over 60% of the produced products are exported, and it is very strongly connected to the European Union; over 75% of its exports are going to the other EU countries. Therefore, the increase of Chinese imports in the EU may have significant adverse effects on the Hungarian producers. Meanwhile, Hungary is a very small player compared to China even on the EU markets. The EU imported products of 375,000 Million USD from China in 2010, while the whole Hungarian export was about a quarter per cent of this value (94 Million USD). Therefore, reverse causality is not likely; the behavior of Hungarian firms hardly influences prices and production volumes in China.

We define our product-level import shock measure as the increase in Chinese imports in the products at the EU markets. The percentage change of imports from China to the EU28 countries from the previous year’s value was calculated on the Prodcom 6-digit level and standardized by the EU28 consumption to correct for effects that may come from the general decline of EU markets in the respective product group: (6) where denotes the percentage change in the ratio of Chinese imports to total consumption in the EU for product p in year t. Note, that by this measurement of import shocks based on product-level imports we also avoid mixing up the increased product competition effect of Chinese imports with their positive effect of making the intermediate goods cheaper, which we should have been addressed if choosing other import shock measure, such as tariff levels.

For our analysis, we calculate the firm-level exposure to increasing import shocks from China by the weighted averages of the import shocks of the different products in the firm’s portfolio: (7) where Y_ipt denotes the sales of firm i of product p in year t, whereas m denotes the number of products the firm produces. Thus, Y_ipt is used as weights in calculating the weighted average.

3.5. Statistical methods

We apply three methods to analyze the relationship between demand shocks and related diversification. In the first exercise, we look at how coherence of the product portfolio changes in firms according to their exposure to negative demand shocks measured by shrinking markets and import shocks (Hypothesis 1a). In this model, we also examine the impact of liquidity mechanism: whether a restricted liquidity amplifies the impact of the demand shock on the coherence of the product portfolio for firms more exposed to financial markets (firms with higher short-term debt) (Hypothesis 1b). Next, we analyze the events of dropping products from the portfolio, and check, whether relatedness is a stronger predictor of keeping a product conditionally to the exposure to negative demand shocks (Hypothesis 2). At last, we analyze the events of adding new products under identical conditions (Hypothesis 3).

In the first approach, we use firm-level panel regressions with firm fixed effects. The firm fixed effects enable us to exclude alternative explanations based on unobserved heterogeneity of firms (e.g. those industries that exhibit more related diversification were also hit harder by shrinking markets or import shocks), as we compare the coherence of the product portfolio in the firm in the year following the identified demand shock with its previous values. Our explanatory variables are the exposure to shrinking markets, or to import shocks. We also add time dummies (D_t) to control for yearly fluctuations in the trends. To assess the impact of financial exposure of the firms, we also add the Leverage ratio (short term debt by equity) to the model, together with its interaction with our main independent variables, Shrinkage exposure and ImpShock exposure to the market shrinkage model. We estimate: (8) and (9) where D_t are year dummies, and ξ_i and ε_it are error terms.

The second hypothesis is tested by examining the likelihood of dropping products conditional to their relatedness to the company’s core product. Our dependent variable is the dummy of dropping a non-core (p≠c) product from the portfolio of a firm (i) in year t, Drop_i,p≠c,t. The key independent variable is the value of relatedness RR_cp,t−1 of the examined product to the core product of the firm, one year prior to the decision. Our specific interest is whether the effect of relatedness is stronger when demand shocks are experienced than in other instances; thus, we include the interaction of the relatedness value with the Market shrinkage and Import shock terms. As controls, we add the main effect of the Market shrinkage and Import shocks, the Leverage ratio and its interactions to the Market shrinkage variable, and year dummies (D_t) to both models: (10)

And for product-level import shocks: (11)

We use fixed-effect panel regressions with the fixed-effects on firm-core product levels to compare only those instances within firms when they did not change their core products. ξ_ic is the corresponding error term. For simplicity, we used linear probability models to estimate Eqs 10 and 11. Note that the units of analysis are firm-products; therefore, our identification comes from comparing products that the firm produces in the given year and in other years.

The estimations defined in Eqs 10–11 may raise concerns of endogeneity. Namely, we predict firms’ decisions on their product portfolio with the value of relatedness that we observe in the portfolio, which in turn comes from diversification decisions of firms, including the focal firm as well. Once again, we can address this issue by lagging relatedness with one year. Thus, the indicator will capture the tendency that the given products have been produced jointly in all firms of our sample with the core product of the firm that we investigate in the regression. This “tendency” certainly includes the previous activity of the examined firm as well. However, we compare products, which were both included in the product portfolio in the previous year. Thus, the difference between their relatedness to the core product must come from the activity of other firms on the market. In sum, we argue that lagging relatedness solves the direct endogeneity issue.

To test diversification into new products, we estimate the relation between relatedness and adding products in situations according to firms’ exposure to demand shocks. The dependent variable is Add_ipt that takes the value of 1 if the product was added to the portfolio of the firm and zero otherwise: (12)

Thus, each product, that were not part of the firm’s product portfolio in the previous year, must be considered for each year and for each firm. For reasons of simplicity, we exclude those products that were not introduced by any firms in the examined year.

To avoid mechanical correlation, we modify the Shrinkage exposure and ImpShock exposure variables for this model by excluding the newly added product from the firm-level average, thus considering only those products that were already in the firm’s portfolio: (13) (14)

Our key predictors are the lagged value of relatedness of each considered product to the core product (RR_cp,t−1), and these adjusted market shrinkage and import shock exposure variables. The hypothesis again considers the interaction terms of these variables. We add year dummies (D_t) to control for yearly fluctuations, and fixed effects for the core product of the firms (ξ_c). In addition, we add the leverage ratio and its interactions to the market shrinkage model. The estimated models are: (15) (16)

4.1. Descriptive statistics

Descriptive statistics of our measures are displayed in Table 2. These distributions indicate that while the revealed relatedness between two randomly selected industries is rather low, coherence of actual firms’ portfolios is much higher. The mean of the Market shrinkage measure is negative, indicating that firms on average enjoyed prosperity during the period, however, its high standard deviation signifies that many of them experienced negative demand shocks in different years. On the other hand, increasing Chinese imports was observed on average throughout the observed period, therefore the mean Import shock value is positive. It is also visible that the Market shrinkage measure is one magnitude lower than the Import shock measure, which is because Hungarian production data is more volatile than Chinese imports in the EU. Furthermore, we observe the coherence measure for fewer firms compared to Market shrinkage and Import shocks, because it can only be calculated for multi-product firms.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Descriptive statistics.

https://doi.org/10.1371/journal.pone.0280461.t002

From the yearly averages illustrated in Fig 4 it is visible that the Market shrinkage follows a very different trend from the Import shock. Market shrinkage increased after Hungary’s joining the European Union in 2004, then decreased in subsequent years. We see an increased variance in 2008, then the mean value of the indicator jumped to its peak in 2009, and it was decreasing in the subsequent years. On the contrary, Chinese imports slowly increased in the EU, except in 2009, when we see a negative spike in the trend of the Import shock measure (Chinese exports are missing for 2012 because of translation ambiguity issues between SITC and CPA6 for this year). The Coherence indicator of firm’s product portfolio fluctuates across the years in the period. As Import shock and Market shrinkage show somewhat opposite trends, and they also may impact different firms, we cannot draw clear conclusions from these aggregate descriptive findings on the correlation between the Coherence and these measures.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Evolution of Coherence, Market shrinkage and Import shock measures.

https://doi.org/10.1371/journal.pone.0280461.g004

The decrease of Chinese exports to the EU countries during the financial crisis of 2008-2009 was primarily due to the decrease in GDP growth of the EU countries (most importantly its biggest markets in the EU: Germany, Netherlands, UK and Italy) [88]. Despite Hungary’s economy is heavily linked to these countries, on the product level we do not observe significant correlations between the Import shock and Market shrinkage measures. We checked this in cross-section, in a pooled (OLS) model, and in a panel model with year FEs using different lags.

4.2. Firm coherence and crisis

Results indicate an increasing coherence of the portfolio after a firm was exposed to import shocks (Table 3), which supports Hypothesis 1a, but no such direct effect can be observed for being exposed to market shrinkage. The positive and significant interaction term of the Shrinkage exposure with the Leverage ratio confirms Hypothesis 1b. The insignificant coefficient of the Shrinkage exposure variable together with its positive and significant interaction term with the Leverage ratio suggests that when experiencing shrinking markets, those firms increase product portfolio coherence, which also have significant financial exposure (high Leverage). When ImpShock exposure is used as the main explanatory variable to explain next year’s Coherence, we find that moving up 1 standard deviation in the shock variable is associated with an increase of 0.021×0.150=0.0032, i.e. 0.015 standard deviation in the Coherence of the firm’s product portfolio in the next year (the coherence of previous year is only a technical variable in the model to serve as reference point for the next year and to capture the stability of the process.). Checking the effect of the leverage ratio and its interaction with the ImpShock exposure variable, we observe that they are not significant and do not alter the positive effect of ImpShock exposure. Thus, the demand shock from import pressures is not moderated by the tightening of financial resources (Table 3 Column 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Change of the Coherence of firm’s portfolios.

https://doi.org/10.1371/journal.pone.0280461.t003

In sum, these results inform us that technological relatedness is a major factor of product portfolio diversification in firms that already have a specialized and technologically coherent product portfolio. These companies follow a strategy that exploits economies of scope. During a financial crisis, negative demand shocks foster specialization and portfolio coherence even further only for firms having financial liabilities. This suggests that companies tend to streamline their production in crisis times probably motivated by cost concerns.

The effect of exposure to market shrinkage or to import shock on the coherence of the firm’s product portfolio may take more years to manifest. Therefore, we test if adding lags of 2 and 3 years of the shock variables change the overall picture of effects, i.e. if we see a dynamically changing effect over years (Table 4). In these regressions, we can only use those firms that have survived for at least four years, which means that our previous sample is cut to half. The fixed-effects regression on the coherence of the next year’s product portfolio are consistent with the baseline specifications showing that exposure to market shrinkage is only significant with its interaction to the leverage ratio. However, Shrinkage exposure and ImpShock exposure show different dynamics in their effect on the product portfolio. In case of Shrinkage exposure, a considerable part of the effect is maintained for the 2^nd year, whereas in case of ImpShock exposure, the effect rather falls back dramatically for the 2^nd year. No significant effects can be seen in the 3^rd year.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Dynamics of the Coherence of firm’s portfolios.

https://doi.org/10.1371/journal.pone.0280461.t004

4.3. Narrowing and expanding the product portfolio

Regression results in which we investigate the events of termination of products from company portfolios are displayed in Table 5. We test two models, both of which include the Relatedness of the product to the core product of the company, the shock variables and their interaction with the Relatedness indicator are introduced separately. In case of the shock associated with decreasing markets, the variables are also interacted with the Leverage ratio.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Dropping products from firm’s portfolios.

https://doi.org/10.1371/journal.pone.0280461.t005

We find that the coefficient of relatedness of the product to the core product of the firm is significant in both cases. Its direction corresponds to the expectation that firms tend to drop those products with higher likelihood, which are less related to their core activity. A further finding suggests that firms are more likely to drop products, which were exposed to Import shock but there is no such effect of Market shrinkage. The interaction term of Relatedness and negative demand shock variables inform us whether firms are more likely to drop their products unrelated to their core product when experiencing negative shocks compared to other cases. This interaction term is insignificant in case of the Market shrinkage variable (Column 1), but it turns negative and significant in case of the Import shock variable. This latter result indicates that firms are more likely to drop an unrelated product in case Chinese import to EU intensifies in this given product. Considering the interaction of the Market shrinkage variable with the Leverage ratio, we see a negative coefficient, suggesting that more indebted firms are more likely to drop any declining demand products. The non-significant triple interaction with the relatedness variable however indicates that these products do not tend to be the more unrelated ones. That is, in times of financial crisis, companies drop products from their portfolios that are more exposed to the contraction of financial resources, but not unrelated products: they drop products for which customer demand is more likely to fall in response to the contraction of financial resources. However, related diversification of firms motivated by economies of scope are intensified by increasing competition from import. In sum, the import shock approach supports Hypothesis 2, but the market shrinkage approach does not.

Table 6 reports results of regressions, in which adding new products is the dependent variable. The estimation strategy is identical with the one from the previous table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. The effect of crisis on adding new products.

https://doi.org/10.1371/journal.pone.0280461.t006

The positive and significant coefficient of Relatedness suggests that firms do not pick new products randomly but add those ones, which are related to their core business. This finding is in line with previous studies on firm diversification [8, 9]. However, we find very small influence of the negative demand shock variables.

Falling demand in the markets of its existing products, that we measure by Shrinkage exposure, does not seem to deter firms from expanding their portfolios, while experiencing import shocks only does so to a minor extent. The interaction term between Shrinkage exposure and Relatedness is positive and significant and informs us that technological relatedness plays an important role here. In the case of falling demand in existing products, firms are likely to diversify into new products that are technologically related to their core product. This result supports Hypothesis 3, and appears to be independent of the indebtedness ratio of the firms, as all the interactions with Leverage ratio are insignificant.

The ImpShock exposure variable and its interaction term with Relatedness behaves differently. We find a positive and significant main effect of import shock and a negative significant interaction. Consequently, Hungarian firms tend to diversify to new unrelated products when their production must face an increasing import competition from China in the European markets. This result contradicts with the expectation formulated in Hypothesis 3 and indicates that firms follow a strategy of risk reduction that might help them to mitigate the increasing competition in their existing markets.