Determinants of Vertical Integration in Poultry Production in Ghana: Application of Count Data Models

Poultry production has significant potential to reduce protein deficiency, food insecurity and poverty in Ghana. However, limited vertical integration and high cost of production in the sector have stifled growth and exposed poultry farms in the country to many risks, leading to poor business performance. This study uses cross-sectional data from 102 commercial poultry farms to assess the determinants of extent of vertical integration in the Ghanaian poultry industry by employing zero-inflated Poisson (ZIP) and Zero-inflated Binomial (ZINB) models. The results show that one in every four poultry farms in the country are vertically integrated, either partially or fully. The ZINB model, which best fits the data, reveals that the extent of vertical integration in the poultry business is significantly influenced by a set of personal (education, occupation, and farming experience) and farm level (land tenure, flock size, production cost, and farm revenue) characteristics as well as institutional factors (credit access, extension access and membership of association). The paper discusses the implications of these findings and provides appropriate recommendations for strengthening the poultry industry in Ghana.


INTRODUCTION
Poultry, widely termed as the "cow" of the poor, has the potential to improve nutritional security and ensure poverty reduction across sub-Saharan Africa (SSA) (Food and Agriculture Organisation, 2010). In Ghana, poultry production makes a significant contribution to the economic growth of the country (Adei and Asante, 2012; Atuahene et al., 2010). The sector accounts for about 34% of domestic meat production and employs nearly 2.5 million people, of which the majority are women who subsist on poultry and other related products for livelihoods (Guèye, 2000;Ministry of Food and Agriculture [MoFA], 2020). Despite its influential contribution to the growth of the economy, the Ghanaian poultry industry, over the past decades, has declined due to intense competition from imported poultry products and high cost of production (Kusi et al., 2015;Anang et al., 2013;Atuahene et al., 2010). For instance, official data shows that domestic broiler meat supply has declined from 60% to 20% culminating in an increase in imports from 13,900MT to over 155,000MT in 2018 ( Figure 1) (FAOSTAT, 2019).
To help create a more competitive and efficient poultry industry, the United States Development Agency (USDA) implemented the Ghana Poultry Project (GPP) from 2015 to 2020 (USDA, 2017). This was further strengthened by the introduction of the "Rearing for Food and Jobs" program by the Government of Ghana to produce 40,000MT of broiler meat by the end of 2020 (MoFA, 2019).
In these initiatives, much emphasis were placed on strategies such as the provision of subsidized inputs, producer's capacity building, and strengthening of buyer-supply linkage to minimize the cost of production and improve the overall competitiveness of the Ghanaian poultry industry (Anang, 2013; Global Agricultural Information Network (GAIN) report, 2017). However, an important management strategy that has a significant influence on the overall performance and competitiveness of the poultry industry but has received little attention in these initiatives is vertical integration. Reasons such as limited empirical data on the implications of vertical integration in poultry farming are adduced to this apparent lack of consideration in poultry development programs of Ghana (Atuahene, 2010). This study, therefore, presents an empirical analysis of the extent and determinants of vertical integration of poultry production in Ghana. A thorough understanding of the implication of vertical integration in poultry farming is a prerequisite to guide policy intervention that will improve the efficiency and competitiveness of the poultry sub-sector of the country.
Empirical studies on vertical integration as a key catalytic operation to expand and improve the competitiveness of firms have been well-documented (Baum, 1951;Coarse, 1937;Buzzel, 1983;Maddigan, 1981;Grega, 2003;Martinez, 2012). Isaksen et al. (2002) suggests that vertical integration of firms helps to reduce production costs while improving efficiencies. In a multi-study of agricultural products, Maritinez (2012) observes that not only does vertical integration enable Year PRODUCTION IMPORTS firms to eliminate risk and uncertainties, but to overcome market failures and expand market power for competitive advantage. Thus, the general motive for firms to integrate vertically is to reduce the overall cost of production, which in turn, improves firms' performance and consumer welfare (Carlton and Perloff, 2005). Despite this, there are only few studies (e.g. Bamiro et al., 2006;Bamiro and Shittu, 2009;Bamiro et al., 2014) across sub-Saharan Africa that consider vertical integration of poultry farms as a conduit to increase competitiveness and efficiency of the sector.
Moreover, the findings of these limited studies have generally been mixed and inconclusive (Bamiro et al., 2006;Bamiro and Shittu, 2009;Bamiro et al., 2014). This is because the overreliance on the use of value-added ratio as a proxy to measure the extent of vertical integration of poultry farms is fairly limited (Barrera-Ray, 1995). According to Hamdaoui and Boyuad (2019), the two economic variables of sales and purchases, which are key indicators of value-added ratio in the determination of vertical integration, may be influenced by other factors (such as production techniques and staff competencies) besides vertical integration. In this study, therefore, we compute a vertical integration index based on available data to measure the extent of vertical integration in poultry production. We also examine the key determinants of extent of vertical integration by paying special attention to important farmer, farm level, and institutional factors.
The rest of the study is organized as follows. First, theoretical concept of vertical integration and its measurement are reviewed in Section II. The research method is presented in Section III before results and discussions in Section IV. Conclusions and recommendations of the study are also outlined in Section V.

VERTICAL INTEGRATION AND ITS MEASUREMENT
The concept of vertical integration has been popular in economics literature since the era of Adams Smith, and the division of labour theory propounded by Young (1928), and Stigler (1951). Yet to date, there is no universally accepted definition of the concept (see, for instance, Coase, 1937;Buzzell 1983;Rehber, 1998;Luciana, 2008;Basant and Mishra, 2017). Despite the diversity in the definitions, a common understanding as adopted in this study suggests that firms are vertically integrated when they partially or wholly internalised their operations without the involvement of external agents. Thus, in a vertically integrated firm, two or more production stages occur under one management where all upstream production activities serve as inputs for downstream activities and vice versa (Hamdaoui and Bouayad, 2019). As a result, the product developed is not transmitted via the market and, hence, does not reflect market prices (Hamdaoui and Bouayad, 2019). In summary, Barrera-Ray (1995) contends that the stages of production in a vertically integrated firm should be contiguous without intermediaries and no market exchanges. There are two basic types of vertical integration: backward and forward vertical integration (Grega, 2003;Barrera-Ray, 1995;Bamiro and Shittu, 2009). A firm engages in backward integration when it produces its input instead of relying on external stakeholders. In the case of forward integration, firms take ownership of upstream activities that include distribution, processing, or supply of the firm's final product to consumers. Therefore, to accurately measure the full implication of vertical integration on a firm, both backward and forward integrations have to be sufficiently captured.
The measurement of vertical integration across industries is complicated and poses several practical and theoretical hurdles, which limit the ability of researchers to examine the extent of vertical integration on firms' performances (Hamdaoui and Bouayad, 2019). Nonetheless, two distinct measures of vertical integration can be identified in the literature. These include measures determined from financial statements and the use of multidimensional constructs such as computation of indices based on available data (Kaiser and Obermaier, 2020).
In terms of financial measures, the Value Added to Sales (VAS), proposed by Adelman (1955), is the most widely used approach to proxy a firm's degree of vertical integration. The VAS is mathematically friendly and has a strong theoretical foundation because it is defined by two economic variables as Buzzel, 1983). Despite its simplicity, the VAS has many drawbacks, which makes it near impossible to be applied in firms that operate in the informal sector such as poultry production in developing economies. First, the approach measures monetary values, which can be influenced by other factors such as efficiencies of production techniques and employees (Hamdaoui & Bouayad, 2019), and not on physical activities/transactions that contribute to the degree of vertical integration (Barrera-Ray, 1995).
Similarly, the measure is criticised as not being symmetric concerning production stages, as it favours upstream activities (Barrera-Ray, 1995;Isaksen et al. 2002). Lastly, not only is the VAS dependent on financial indicators that are sensitive and confidential, but records on these indicators are poorly kept, especially for informal firms in developing countries (Essel et al., 2019). However, the VAS is the dominant approach used in the few existing studies that consider vertical integration of poultry production across sub-Saharan Africa (See, for instance, Bamiro et al., 2006;Bamiro and Shittu, 2009;Bamiro et al., 2012).
In this study, we adopt a more data-driven approach (vertical integration indices) that permits the use of reliable and readily available data of poultry farms to measure the extent of vertical integration. The indices proposed by Chapman and Ashton (1914), and Gort (1962) based on the number of equipment and employees respectively used in different stages of production within the firm were adopted and modified to calculate the extent of vertical integration of the Ghanaian poultry industry. In this modified approach, the poultry farm's main activity (i.e., production of eggs and meat) is separated from its auxiliary activities and values assigned to each activity in the poultry value chain. Empirically, six (6) major auxiliary activities are performed along the value chain. These include ownership of crop farm (mainly maize), feed mill, hatchery, delivery van, processing plant and retail outlet (Begum, 2005). The number of activities engaged in by each poultry farm is expressed as a ratio to the number of major activities in the value chain.
Mathematically, the degree of vertical integration adopted in this study is expressed as: where, is the extent of vertical integration expressed in percentage, is the number of activities engaged in by ℎ poultry farm and represents the number of major auxiliary production stages in the poultry value chain.
This approach is similar to the index employed by Hamdaoui and Bouayad (2019) to measure the extent of vertical integration in the Moroccan textile industry. The following criteria as defined by Misund (2016) are used to categorise the poultry farms based on the extent of vertical integration, which was used in the further econometric analysis (see Table 1).

Table 1 Benchmark for the Levels of integration Ratio (Percentage) Level of vertical integration Less than 20%
Non-integrated 20% to 65% Partially integrated Greater than 65% Fully integrated Source: Adopted from Misund (2016)

Study area
The study was conducted in the Dormaa municipality located in the western part of the Bono region of Ghana ( Figure 1). The municipality has a total land area of 1210.28km 2 with a population of 112,111, representing 4.9% of the national population. Dormaa has an agrarian economy that employs 68.4% of its total population. About 73% of the population is located in rural communities. Crop and livestock framings are the major agricultural activities in the area. The Municipality is noted for the production of poultry, which constitutes more than 70% of the total livestock production in the area. Besides, the old Brong Ahafo region (now comprising Bono East, Bono, and Ahafo) is ranked third in terms of poultry production in Ghana with Dormaa municipality contributing more than 80% to the regions' poultry population (Anang et al., 2013;Kusi et al., 2015). Though there are two major poultry production lines; broiler and layer, nearly

Study design and data collection procedure
The mixed research technique that comprises both quantitative and qualitative designs was used to carry out the study. Combining the two techniques helps to improve the reliability and validity of the data collected through data triangulation. Qualitative methods such as focus group discussions and key informant interviews were conducted with leaders of the poultry associations and the Municipal agricultural officers. The focus group discussion comprises seven (7) participants consisting of four (4) male and three (3) female poultry farm owners. In total, 4 focus group discussions were conducted with one each in the 4 selected poultry producing communities in the study area. Key qualitative data collected included farmers' perception of vertical integration, bottlenecks to practicing vertical integration as well as general production and marketing information. On the other hand, a structured questionnaire was developed and used to solicit quantitative information on the poultry farms (i.e., size of farms, production cost, revenue, upstream and downstream poultry activities, among others), producers' demographics, and access to relevant institutions such as the veterinary services.
The study employs cross-sectional data collected between February and March 2020. The data collected was based on the 2019 production year. Prior to the data collection, the survey questionnaire was pre-tested in one community in the study area to assess the appropriateness of the statements towards meeting the objectives of the study. Seven (7) poultry farmers were randomly selected and used in the pre-testing.
The study conducted a full census to include all the 137 registered commercial poultry farms in the Dormaa municipality. The list of all the farms was obtained from the leadership of the poultry farmers' associations in the municipality. The veracity of the list was authenticated at the Municipal Department of Agriculture. Information on other poultry farms not registered were also solicited from the Department and were contacted for data collection. In total, managers and owners of 102 poultry farms were available for data collection within the survey period. The communities included in the survey are Wamfie (15) 1 , Dormaa Ahenkro (35), Kyeremasu (12), Asuochia (10), Nkrankwanta (30), and Nsesreso (10).

Analytical approach
Descriptive tools including frequency tables, pie chart and measures of central tendencies and dispersions were used to summarise key farm level and personal characteristics. The zero-inflated Poisson (ZIP) and negative binomial (ZINB) regression models were used to examine the precursors of extent of vertical integration in the poultry business. The zero-inflated models were chosen for the study because less than half of all poultry farms in Dormaa were found to be vertically integrated; leading to the situation/problem of excess zeros in terms of extent of integration (Atuahene et al., 2010). The ZIP and ZINB models were compared and the model that best fitted the data was selected for further discussion.

Zero-inflated Poisson and Negative Binomial models
In socio-economic studies, outcomes of interest are sometimes count data with excessive zeros (Fang, 2013). While these zeros are important and meaningful, most researchers often treat them as missing values or delete them. In other cases, the data is either transformed into a linear model (which violates the normality assumption) or coded as a categorical dummy variable where all zeros are considered as 'absent' and those observed as 'present' (Lewsey and Thomson, 2004;Yusuf et al. 2018). Under such circumstances, the analysis becomes less useful and less informative if the interest is to determine the number of occurrences (Yusuf et al., 2018). Zero-inflated model can distinguish between the two processes causing the excess zeros (Diallo et al., 2019;Yusuf et al. 2018). A common feature of zero-inflated model is its ability to simultaneously produce two outcomes in count data models by: i.) examining the effects of covariates on the extra/inflated zeros and, ii) generating Poisson or negative binomial aspect of the model (Diallo et al., 2019;Fang, 2013;Lewsey and Thomson, 2004).
Zero-inflated Poisson and -negative binomial models are specialized types of Poisson regression models that are widely employed in count data analyses with inflated zeros (Fang, 2013;Lambert 1992;Diop et al., 2016;Diallo et al., 2017). Lambert (1992) first developed the zero-inflated Poisson after the standard Poisson regression model failed to produce efficient estimates with excess zeros in count data variables. Similarly, modeling a zero-inflated count data that has overdispersion problems with ZIP also produces coefficients that are consistent but inefficient (Fameye et al., 2003;Fang, 2013). Greene (2003) therefore, proposed the use of ZINB to account for the over-dispersion problem under such circumstances. Over-dispersion in count data models arises when the variance of the scaler-dependent variable is larger than its mean (Winkelmann and Zimmermann, 1995;Winkelmann and Zimmermann, 1998).
In the ZIP model, the scaler dependent variable ( 1 , 2 … … … … … … … … … … … … … ) is independent and the assumption behind the model is that given a probability( ), there are two possible outcomes; 0 and the probability of (1 − ) which leads to the generation of a Poisson random variable (λ) in (Cameron and Trivedi, 2013). The distribution of is given as follows: The variance and mean of the zero inflated Poisson distribution are specified in (2) and (3), respectively; Similar to ZIP, the ZINB also has two possible outcomes. Assume π as the probability for the occurrence of case 1 which is zero and (1 − ) as the probability for case 2 which is a success. If (1 − ) occurs, the counts (including zeros) generated are line negative binomial model. In this case, Greene (1994) defined the probability of the ZINB random variable, as; This implies that, From (5), the mean and variance of becomes: where λ denotes the mean of the negative binomial distribution with being the over-dispersion parameter. As → 0, the ZINB distributions reduces to the ZIP. Meanwhile, λ is expressed as a function of linear predictor: where is a vector of unknown parameters to be estimated from the covariate vector ′ that would include farm and non-farm related factors that influence the extent of vertical integration of poultry farms. The main estimation procedure for (6) is using the method of maximum likelihood. As noted earlier, both ZIP and ZINB generate two models; first, the count model used to predict the response variable; and second, the inflated model used to predict the occurrence of the excess zeros.

Model comparisons and selection
Three tests of model fits were performed to compare and select the model that best explained the data (Table 1). First, the Akaike Information Criterion (AIC) (Akaike, 1973) and Bayesian Information Criterion (BIC) (Schwarz, 1978) tests were performed to score and select the appropriate model. However, while the AIC is asymptotically efficient but inconsistent, the BIC is consistent but not asymptotically efficient (Cavanaugh and Neath, 2019). In both instances, the model with the smallest value is considered the better fit. The Vuong test was also performed on the two models against the standard Poisson regression and negative binomial models.

Empirical model specifications
Following the theoretical review of both the ZIP and ZINB, the empirical model guiding this study is specified as: hypothesized to contain excess zeros (inflated) and the reasons for such zeros to occur are different from the reasons for a poultry farm to participate in vertical integration. β1……… β17 are the vector of parameters to be estimated, β0 is the constant term, and μi the error term. In Table 3 are the descriptions and a priori expectations of the respondents, farm, and institutional variables included in the model. The explanatory variables adopted in this study were based on the findings from previous studies (Begum. 2005;Issa and Chrysostome, 2015;Harianto et al., 2019) across different agri-businesses in developing countries.

Extent of vertical integration in poultry business
The extent of vertical integration is measured after taking the ratio of the poultry farm's auxiliary activities (besides the core production stage) to the total number of activities along the value chain (Table 4). The ratio is expressed in percentages (Figure 2) to depict the extent to which the poultry farms are vertically integrated. Out of the six (6) major auxiliary poultry value chain activities, 27 of the farms representing 64.3% own and operate their feed mills for mixing feeds. Similarly, about 54.8% owned delivery vans for both wholesale and retail delivery of eggs and chicken carcass within and outside the study's region. Besides, 42.9% of the respondents possess retail outlets in urban consuming centers to dispose of their eggs and birds directly to consumers. The data further shows a significant number (38.1%) of the poultry firms managing their maize farms; the major feed ingredient representing 60% of compound feeds (USDA, 2017) used for both layers and broilers in the study zone. However, there were only one (2.4%) and two (4.8%) farms that have hatchery and processing houses, respectively. The absence of hatcheries to breed local day-old chicks is not uncommon since most poultry farms in Ghana prefer foreign day-old chicks from Europe compared with domestically hatched day-old chicks. According to Boschloo (2020), dayold chicks from Europe are hardy, disease-resistant, and could recover quickly after sickness compared with the domestically hatched chicks that are generally of low quality. In support, the GPP reported that more than 511,960 broiler and 7,130,999 layer day-old chicks are imported into Ghana on annual basis (USDA, 2017).  Figure 2 shows the levels of vertical integration based on the classification by Misund (2016).
Nearly three-quarters (74%) of the surveyed poultry farms fall below 20% of vertical integration and are classified as non-integrated farms. Partially integrated farms (21% and 65% of VI) represent 22% while fully vertically integrated farms are less than 5% in the study area. This finding agrees perfectly with the observations made by Bamiro and Shittu (2009) who reported significant non-integrated farms, but few full and partially vertically integrated poultry farms in Nigeria. The low degree of integration for the poultry farms may have a negative implication on the cost of production since farmers are likely to depend on intermediaries to source inputs (feeds, day-old chicks) and to dispose of the final products (egg and broiler meats). According to Begum (2005), high transaction and searching costs contribute to increasing the overall costs of producing poultry in developing economies.

Variable description according to extent of vertical integration
Male farmers operate the majority (69.6%) of the poultry farms, which is slightly lower than the 89.5% reported by Adei and Asante (2012) in the same study municipality (Table 5). The low proportion of females in the poultry business may be attributed to the socio-cultural and economic constraints faced by women in establishing business ventures in developing economies (McPherson, 1992;Robinson & Sexton, 1994;Presser & Baldwin, 1980). The capital demand to set up and maintain poultry farms in sub-Sahara Africa is high, which turns to limit women participation in the livestock business. The high literacy rate of 52.9% of poultry farmers with more than senior high school certificates could have positive implications for the growth of the poultry business. This is because educated farmers can read and write which improves their ability to keep proper farm records, access information/credit, and adopt technologies to increase production. The literacy data is consistent with the 43.4% of poultry farmers with senior high 74% 22% 4% school and tertiary certificates reported by Amoabeng (2011) in the same study area. Likewise, 78.4% of the poultry farmers are full-time workers, which emphasised that poultry farming is a major source of livelihood and thus can serve as a conduit for poverty reduction in the study area.
This finding relates well with Bamiro et al (2009) who reported that over 50% of poultry farmers particularly in West African countries such as Nigeria are full-time workers. The high FBOs membership of 85.3% presupposes that, through the leadership, the members can have access to reliable information and productive resource to improve poultry production/productivity. This data is consistent with the report by Nimoh et al. (2013) (1983) and Bamiro et al. (2009) who conclude that vertical integration leads to cost reduction, which, in turn, increases investors' investments.

Parameter estimates from ZIP and ZINB regression models
The coefficients of both zero-inflated Poisson and zero-inflated negative binomial regressions are summarised in Table 6 and Table 7 Table 8, various tests were computed to compare and select the best model that describes that data. First, the Voung tests for both models are significant at 1% significance level, which implies that the data perfectly fits ZIP and ZINB due to the excess zeros instead of the standard Poisson and negative binomial models, respectively. However, the sample mean (

Personal characteristics
The

Farm characteristics
The type of land ownership tends to significantly influence the extent of vertical integration of poultry farms in the study area. For instance, the extent of vertical integration is expected to increase exp (0.255) = 1.29 times for farmers with full property rights of farmland than farmers with family/inherited farmlands. This finding supports the assertions made by Awudulai et al. Chrysostome (2015) who documented a significant positive relationship between farm size and the capacity to vertically integrate agribusinesses in Rwanda. Likewise, Elzo (2010) asserts that agribusinesses with higher revenue tend to record higher profitability and as such, such businesses will have enough funds for investments in other activities that increase overall firm performance.
However, the data shows that increases in overall production costs are expected to decrease the extent of vertical integration by a factor of exp (-0.0290) = 0.97, all things being equal. This finding according to Kusi et al. (2015) partly explains the low vertical integration among poultry farms in Ghana. This is so because the high cost of production leads to low profitability of the poultry business, which eventually generates little or no extra funds to invest in other activities along the poultry value chain.

Institutional characteristics
Access to institutional factors such as credit facilities, extension contact, and membership of poultry farm association are well recognized to create the enabling environment for investment and expansion of existing businesses (Essel et al., 2018). The data shows that the extent of vertical integration for farmers with access to credit facilities is expected to increase by exp (0.354) = 1.42 times compared to farmers without credit access, all things being equal. This finding corroborates with de-Janvry et al. (2005) who noted that access to credit/loan improves the liquidity capacity of the farm; helps smoothen capital fluctuations, and thus facilitates investments in other activities that improve overall business performance. In terms of extension contact and membership of poultry association, the results depict that the extent of vertical integration for farmers with extension contact and membership of poultry association is expected to increase by exp (0.2612) = 1.26 times and exp (0.2980) = 1.37 times, respectively, all things being equal. These findings did not deviate from the observations made by Marinda et al. (2006) who reported that the production and marketing landscape of agricultural products is evolving fast, and this requires the collection and processing of information to gain competitive advantage and expand on-farm investments. Thus, farmers with improved extension service contact and membership of associations tend to be abreast with improved farming technologies and can access credit facilities for more farm investments to achieve higher profitability.

The logit inflation model
The inflation component of the ZINB predicts the occurrence of the excess zeros of the model (Table 7). The data shows that farmer personal factors such as education, primary occupation, and However, an increase in farming experience is likely to increase the odds of being in the absolute zero categories by exp (0.5394) = 1.71.
In terms of poultry farm-related factors, whiles the odds of a certain zero is lower for farms with higher flock size, employee size, and revenue, the odds are higher for farms with a high cost of production. The results also show a higher odds ratio for farmers with full outright ownership of land compared to family/inheritance ownership, all things being equal. The result implies that increasing flock size, employee size, revenue with full outright land ownership contribute less to being part of the absolute zeros in assessing vertical integration in poultry production. However, a higher cost of production predisposes farmers to belong to the excess zero categories.
The study shows two institutional factors including credit access and association membership significantly influence the absolute zeros of vertical integration. The data shows a lower odds ratio for farmers with credit access to be part of the absolute zeros categories in examining vertical integration of poultry production. On the contrary, access to association membership tends to increase the odds of poultry farmers belonging to the absolute zero groups.

CONCLUSION
Over the past decades, the poultry industry in sub-Sahara Africa has declined due to the high cost of production. Strategies that enhance the vertical integration of poultry farms would greatly influence transaction costs, risks, and uncertainties as well as demand variations, which ultimately improves the competitiveness of the sector for higher farmer returns. However, little is known about the implications of vertical integration in the poultry sector, particularly in Ghana. This study, therefore, examines vertical integration in poultry production using econometric models that provide findings with serious implications for the development of the poultry industry. The study contributes to existing agribusiness management literature by exploring critical factors that influence the vertical integration of poultry farms, particularly in Ghana.
Given that previous studies on the measurement of vertical integration in poultry production are simplistic and inconclusive, this study uses vertical integration index to accurately and sufficiently capture the extent of vertical integration in the industry. The study evidences that institutional factors such as membership of poultry associations, extension education, and access to credit are important precursors of vertical integration among poultry farms. This finding has implications to strengthen existing poultry associations through periodic capacity building programs for both leadership and members. This is even more important because the study shows a significant relationship between farmer's characteristics such as formal education and the decision to participate in the vertical integration of poultry farms. To complement this effort, special concessionary credit facilities could be made available to members of these associations for diversification of investments along the poultry value chain. Second, the significant effect of farm factors such as costs of production on vertical integration of poultry business demands subsidy or elimination of import duties on critical poultry inputs such as day-old chicks and medications into the country. Lastly, the study shows that the ZINB model best describes the determinant of vertical integration for data with excess zeros and over-dispersion. Therefore, it is highly recommended to use objective criteria in choosing appropriate econometric models to analyse count data problems that are zero-inflated and over-dispersed.

Availability of data and material
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.