Productivity Spillovers From Foreign Direct Investment To Ethiopian Manufacturing Sector

The study attempted to investigate productivity spillover from FDI to domestic firms in Ethiopian manufacturing industries. The analysis was conducted using panel data for the years 2011 up to 2016 on 260 large and medium scale-manufacturing firms deployed under 24 industries. The System Generalized Method of Moments (SYS-GMM) estimator was employed to conduct the analysis. Accordingly, the study found that a coexistence of both negative and positive effects from FDI through productivity spillover to domestically owned firms using absorptive capacity as determining factor of spillover occurrence in moderate level. Specifically, foreign presence in industry contributed a positive and significant horizontal and backward productivity spillover effect to domestic firms within the industry in average level. The horizontal productivity spillover transmitted to local firms through demonstration and competition effects in moderate level. Likewise, vertical productivity spillover occurred through the channel of sales of intermediate goods and services to foreign firms. Finally, we can conclude that technology gap is critical factor among those factors that determines the productivity spillover occurrence. study. The possible mechanisms through which the horizontal spillover effect a positive and statistically significant coefficient at 1% level. This shows FDI presence benefits depend on the industry absorptive capacity.


Back Ground of the Study
Foreign direct investment (FDI) has outlined as future investment that involves injection of foreign funds into associate enterprise that operates in a very different country of origin from the capitalist. During the last 20 years', the recognition on the importance of foreign direct investment (FDI) within the developing policy and strategies of the emerging economy was growing. To this end, substantial range of developing countries in transition became a lot of receptive to FDI and is exploring ways in which for increasing inflows (Miroux et.al, 2009).
In addition, this development may be through spillover effects whose presence will have an effect on development of business enterprises within the host economy. Theoretically, FDI in developing countries have perceived not solely as a supply of capital flow, however conjointly as a vehicle for feat trendy technology and the necessary managerial skills that these countries need for development. These have number of the explanations why most of the developing countries have continued to pursue domestic policies that encourage many FDI inflows. To realize these, they have provided a number of investment incentive packages, such as lowering income taxes, import duty exemptions, tax holidays and subsides for infrastructures facilities with the motive of attracting more foreign direct investment (Waldkirch and Ofosu, 2010;Glass and Saggi, 2002).
According to Blomstrom et al., (2003), the well-known channels through which FDIinduced externalities to domestic firms are: -(1) demonstration effect, which may help the domestic firms to imitate, (2) competition effect (creating competitive atmosphere between firms), (3) enhancement of export capacity through the reduction of export related costs and the last is backward and forward linkages with foreign firms in the industries.
As a result, since the mid Eighties, the speed of growth of worldwide outflow of FDI has well exceeded that of world GDP, worldwide exports and domestic investment. The developed countries have continued to draw in the majority of the inflows (UNCTAD, 2016). However, recent Proof indicates that the flow of FDI to developing countries has inflated well. UNCTAD (2014) report shows that the share of FDI flows destined to the developing countries reached 54% of the world FDI lows in 2013 compared to 39% for developed countries In contrary, despite efforts to attract foreign direct investment (FDI) by African countries has been undertaken, FDI flows to the continent still decline. Consistent with the UN Conference on Trade and Development (UNCTAD) World Investment Report on UNCTAD (2017), FDI flows to Africa fell by 3 % from $61 billion in 2015 to $59 billion in 2016, however with variance across regions and countries. The general decline of FDI flows to the continent is owing to weak trade goods costs. FDI inflows stay unevenly distributed across the continentwith solely 5 countries (Angola, Egypt, Nigeria, Ghana and Ethiopia) hosting 57 % of continent's total FDI inflows, whereas Africa accounts for 3.4 % share of world FDI. East Africa received $7.1 billion in FDI in 2016, which is greater by 13% from in flow in 2015.This increment was driven by increase in countries, Ethiopia, Mauritius and Madagascar. However, Kenya and Tanzania fell by 36 % to $394 million and 15 % to $1.4 billion respectively. Flows to African nation rose by 46 % to $3. The finding of empirical studies has proven the presence of productivity spillover effect from multinational firms to domestic firms Tomohara & Yokota, 2006;Getachew ,2013;Begum 2016). Despite of the above clear theoretical conception and empirical results that FDI generates positive externalities, there are controversies in the findings of former studies (Yuri ,2007;Konings ,2001;Bruhn & Calegario, 2014, Lemma & Kitaw, 2014.
This contradiction creates a question on the aggressiveness of government's policies action taken to attract in FDI with long run goal which is productivity spillover effect from foreign firm to domestic's firms.
So far as the author knows, no study has been done yet on examining productivity spillover from FDI to domestic manufacturing firms based on theory of sensitivity of domestic firms to receive productivity spillover from FDI firms. This research also differs from the former researches by giving more concern for endogeneity and simultaneity problem. Furthermore, researches have not yet conducted in this topic using firm level dynamic penal data using system GMM estimator modeling in Ethiopia.
Therefore, this study aimed to investigates whether the manufacturing domestic firms in Ethiopia have benefited from foreign firms via productivity spillover or not, and the determinants FDI spillovers occurrence. The study also examines the labor market competition effect resulted by foreign presence on domestic manufacturing firms. The survey data collected from large and medium scale manufacturing sector by Ethiopian central statistical agency in the period between 2011 and 2016 used for analysis.

Theoretical Framework of FDI and Productivity Spillovers
Foreign investment can be a direct investment or portfolio investment. Direct investment is the acquisition or construction of an actual capital by an enterprise from a source country in the host country. Thus, FDI is an investment involving a long-term relationship and controlled by a resident entity of one country, located in a country other than investment country (Duce and España, 2003). Wei and Liu (2001) theorized the presence of multinational firms has possibility to speed up the technology transfer process and reduce the costs of the technology transfer. Competition by multinationals may encourage local firms to innovate and to operate more efficiently. As mentioned earlier, competition, demonstration, learning-by-imitation, contagion effects and the training of workers by foreign firms may help to facilitate the speed of the transfer of technology to domestic firms via labor turnover. Gorg and Grееnаwау (2004) also theorized possible mechanism through which spillover may occur, such as imitation, worker mobility, competition, and linkages. However, theory developed by Gachino (2010) states that Spillover effect occurrence have a strong relationship with the potential of hosting enterprises to absorb know how skills and technologies.
In contrary, Yuri (2007) argued that foreign companies might have negative effects on domestic firms' output and potency if they take over their market or take over their best competent workers.
If the simplest workers leave for foreign companies, potency within the domestic companies declines, that eventually affects the productivity of the domestic companies.

Vertical Productivity Spillovers
Vertical spillovers happen when inter industry spillovers arise mainly by customer -supplier relationship. There exist two types of linkage between the domestic and foreign firms, i.e backward and forward linkage. The backward linkage occurs when the local supplier firms have to meet the demand from the foreign firm in the form of higher quality, price and delivery standards (Smarzynska, 2003). When these local firms supply certain raw materials, the high quality, reliability and speed of delivery that MNCs affiliates demand force them to enhance productivity.
In some cases, local suppliers upstream receive technical and managerial training in the production of required inputs called backward spillover effect. This is likely to generate additional economic activity, income and transfer of technological and managerial skills to the host country

Horizontal Spillovers
According to Görg and Greenway (2004), the domestic firms can benefit through horizontal spillovers through three channels. Firstly, via demonstration effects in which the local enterprises become familiar with superior technologies, marketing and managerial practices used in foreign affiliates. Thus, spillovers can take place in the form of imitating the foreign subsidiaries' technology. The local enterprise may learn simply by observing and imitating the multinationals (Subash, 2006). Secondly, via labor turnover, which occurs when employees from foreign affiliates leave multinationals to join local firms. Through this, knowledge and skills are transferred from the foreign to local enterprises.
The last channel of transmission is competition effect, which occurs when the presence of a foreign firm exerts pressure on local firms to adopt methods that are more efficient. This can allow the domestic firm to survive successfully or even compete with foreign firms. Due to their nature of entry for example efficient management, heavy capital investment among others, foreign firms have an advantage over domestic enterprises (Subash, 2006).

Empirical Evidence on FDI Productivity Spillover
Assessing the effect of productivity spillovers is becoming an essential device to design countrywide regulations regarding inward FDI. So that number of empirical studies has been attempting to measure the value of FDI spillovers, particularly, in the case of developing countries.
Parts of the literatures come up with positive result regarding the FDI productivity spillover effect on the productivity of domestic firms. Contrarily, the other counterpart appeared with nonexistence and even negative results in some developing countries.
To mention some, Tomohara and Yokota (2006) conducted study using plant level data in Thailand and found that FDI affects the domestic firm productivity through horizontal and backward productivity spillover effect on average, while it has no effect through forward linkage. However, challenge full to conclude this because both foreign and domestic firms were used in the analysis.
Consequentially, the results may cause overestimation on the estimated coefficients of parameters.
Similarly, Turi (2015) has done a study to evaluate the productivity spillover from foreign firms to local firms by using unbalanced panel data of large and medium scale-manufacturing industries In other hand, Gасhіnо (2010) conducted a study to identify the determinants of spillover using evidence from west Bengal engineering industries. He found that the occurrence of productivity spillover mainly depends on absorptive capacity, presence of supportive structures and institutions, presence of interactions and trade orientation rather than the presence of foreign firms. In addition, he also includes firm size, age, ownership structure, performance, firm strategy and industry structures. Likewise, Lemma & Kitaw (2014) examined the role of Foreign Direct Investment (FDI) on technology transfer in Ethiopian metal and engineering industries by using survey data of 47 metal and engineering industries. Their findings show that the technology absorptive capacity of local firm is poor. Thus, local firms to adopt, modify and improve a given technology is very weak, un-collaborative operating environment between foreign and local industries.
The analysis was undertaken based on sensitivity of domestic firms to receive productivity spillover from the foreign owned firm so called conditional spillover effect theory. According to this theory, the occurrence of the productivity spillover effect from foreign firms to domestic firms depends on the abortive capacity of the domestic firm, firms' market orientation (import and export), sectoral competition, size of the firm, foreign participation and other related characteristics of a firm or an industry (Schoors and Van der Tol, 2002;Javorcik, 2004;Gachino,2010;Görg and Greenaway, 2004;Girma and Gorg, 2005;Mugendi ,2014;Bruhn and Calegario, 2014).
In other hand, quality of labor which is measured by proxy variable skill intensity gap, ownership ratio, capital intensity gap, embodied technology gap between domestic and foreign firms, organizational capabilities like practice of skilled labor, and experience are factors affecting spillover effect (Kokko et al ,1996;and Blomstrom, 1986). The result indicates that when there is a small technological gap between domestic and foreign firms, the positive effect can be far more success full owing to the higher absorptive capacity.
Similarly, the horizontal and vertical productivity spillover, which are caused by interaction among firms by supplying or purchasing either inputs or finished materials each other are the main interest area of this study. The possible mechanisms through which the horizontal spillover effect may occurs are imitation, worker mobility and competition and vertical spillover effect occurs through linkages (Görg and Greenaway, 2004).
Generally, positive spillover effects depend on a number of factors, including host country's openness to trade, the capacity of its local firms to internalize spillovers, and the ability of sectors to support learning. Another factor is the technological gap between foreign and local firms confirmed by De Mello (1999), as well as the existence of relatively developed local financial markets and qualified human capital (Görg and Greenaway, 2004).

Operational Definition of Variables
The objectives this study which is investigating the productivity spillover from FDI to domestic firms in Ethiopian manufacturing industries is turned into operational definition of variables as follows: i.
Manufacturing industries in this study are defined as large and medium scale manufacturing industries in which at least ten people are deployed as worker, use powered machines, and covers public, private and foreign owned firms in all regions of the country.
ii. Productivity spillover from FDI to domestic firms is defined as the influence of presence of foreign-owned firms on productivity of domestic owned firms through knowledge spread. It is measured by proxy variable labor productivity of firm.
iii. Based on the definition of IMF (1993) and OECD (1996) foreign firms are defined as one in which there is at least 10 per cent foreign equity in the firm.
iv. In contrary, a firm is defined as domestic firms when foreign equity share is less than10 percent.

Data Description
This paper relied on annual survey of large and medium scale manufacturing industries data,

Data Analysis
For the analysis of the data, both descriptive and inferential method has used. From the descriptive statistics, mean, standard deviation, minimum and maximum of the variables, percentage and graphs are used. Heckman (1979) selection bias correction model is also used to test the whether the presence of selection bias. Since the data are a dynamic panel set, system GMM model has employed to control the endogeneity problem. The estimation was facilitated by using STATA version 14, specifically, using STATA's XTABOND2 userwritten command by Roodman (2003) for estimation. One-sample t test was also used to test the mean productivity difference between foreign and domestic owned firms.

Econometric Model Specification
Many studies used an augmented Cobb Douglas production function to overcome the misspecification problem because of it provides efficient estimates (Wang and Schmidt, 2002;De Mello, 1999). Hence, this study used the econometric specification approach used by many scholars (Вuсklеу et al., 2010;Наlе and Lоng, 2011;Nhamo , 2011 ;Xu andSheng(2011), Pham (2016), Negara and Latif , 2012). By using these studies as a base, the econometric model for firm level data analysis has constructed using Arellano and Bond (1991)  Where denotes the mean of total value of firm production per labor in industry level in each year. MVADjt-1denotes mean of lag value of total value of firm production per labor in industry level in each year. Forwardij denotes forward spillover, Backwardij denotes backward spillover and Labcomij denotes labor competition effect. The other remaining explanatory variables denotes the mean of observed variable in equation (1).
Similarly, to meet the objective three, the spillover identified in objective one (horizontal spillover), were regressed against their determinants in addition to technology gap and skill using SYS-GMM estimator based on Green (2006) and Baltagi et al (2015) is presented as follows: Where denotes the mean of horizontal spillover of domestic firm from foreign firms in the same industries. −1 denotes mean of lag value of horizontal spillover of domestic industries from foreign firms in the same industry. denotes a vector of the mean of explanatory variables such as skill intensity gap, technology gap, demonstration effect, labor turn over and competition effects. Skill is represented by average wage of workers. Demonstration effect was represented by percentage of firms who has changed production technique (process) due to competitive pressure (in order to keep up) from FDI within the same industry. Labor turnover is represented by the proxy variable the percentage of firms in the same industry who hired employees with previous experience in FDI firms. Competition effect is proxied by the percentage of firms in the same industry that directly adopted or imitates production techniques by observing or copying from FDI competitors. The measurement of variables and the expected signs of the coefficients of the explanatory variables for the above models were presented in Appendix B for further detail. The difference between the wage of a worker in a domestic firm and the average wage payment of a worker in foreign owned firms centered at its mean multiplied by centered foreign at its mean. + FP*TG The interaction of foreign presence with Technological gap The difference between the logarithm of firm's labor productivity and the logarithm of average labor productivity in foreign firms in the same industry multiplied by foreign presence (all are centered at their means) + Horz Horizontal spillover = * ∑ Where, * ,is foreign firms' output and ∑ is Total output +

Descriptive Result Analysis
In the summary statistics indicates that the employment of capital and raw material inputs in foreign firms were larger than the counterpart.
As presented in the summary  As far as the market concentration is the concern of this study, its value ranges from a minimum of 6.19E-07 to maximum of 0.9153, with mean value of 0.0132 as portrayed in Table 4.1. Since, the average value of the HI is somewhat low, it may signify that on average firms do not have a higher market power. Firm size also ranged from 0.000072 to 144.4533, with mean value of 0.893 suggesting the existence of huge variation among the firm's total value of production in industries.
In the case of continuous dependent variables, the system GMM estimator (GMMSYS) resolves this type of problem.

SYS-GMM Estimation Pre-test
As shown in the correlation matrix table 4.3. below, the firm productivity is positively related with explanatory variables such as capital intensity, labor, material input intensity, foreign presence, herfanlalex index, firm size and horizontal spillover effect. In contrary, it is negatively correlated with technological gap, capital intensity gap and skill intensity gaps between domestic and foreign owned firms. Except the highly strong correction between Herfindahl index and size of firms, the correlation score between other variables does not shows strong correlation. The highly strong correlation between Herfindahl index and size of firm may prove difficult to use them at the time in a single regression because of multicollinearity problem. Accordingly, based on the economic theory, this study uses Herfindahl index only for analysis. Since, Herfindahl index has better explaining power on firm productivity than firm size in relation to the industry and an indicator of the amount of competition among them.  Source: Author's estimation based on data series discussed in methodology

Diagnostic Test
Since only domestic firms were considered in the analysis, the Heckman two-step selection model was estimated to correct the selection bias resulted from the exclusion of foreign firms. The result of estimation was presented in the table 4.4. The specified result shows that the Mill ratio coefficient is not statistically significant in any conventional level of significance. Thus, we do not reject the null hypothesis. This implies that there is no selection bias in the specified model.

. Estimation Heckman two-step Sample
Source: Author's estimation based on data series discussed in methodology As showed in Table 4 ii. The F-test of joint significance reports that the null hypothesis estimated coefficients on the regressors are jointly equal to zero (p=0.000) for the three models is rejected and is satisfied across all estimation for both estimations  Ons |   (2) process is not rejected for the three estimations. This test is therefore supporting the validity of the model specifications.
It is noteworthy that the autocorrelation test and robust estimates of the coefficient standard errors assumes no correlation across individual in the idiosyncratic disturbance (Roodman,2009b (1) in firm level, equation (2) in industry level for FDI productivity spillover analysis and equation (3) in industry level for horizontal spillover occurrence determinants analysis, respectively. As presented in Table 4.6, coefficients for all estimates of the explanatory variables show the expected signs. Most of the explanatory variables such as logarithm of labor, logarithm of capital-labor ratio, logarithm of material input per labor, foreign presence and Horizontal spillover has the expected positive signs and except foreign presence (FP), they are highly significant at 1% level significance. Thus, a one percent increase in the capital stock per worker ceteris paribus raises domestic manufacturing firm's productivity with about 0.09 percent. Likewise, 1% increase in material inputs result in an increase of 0.24 percent domestic firm productivity, ceteris paribus. The coefficient of labor in put also implies that a 1% increase in labor input result in an increase of 0.29% domestic firm productivity, ceteris paribus.
As shown in table 4.6 column 2, the coefficient of foreign presence is positive and statistically significant at the 5% level, suggests that there are productivity gains associated with foreign presence in industry. This might be a sign for the occurrence of direct effect of FDI in domestic firm productivity in the manufacturing sector. The coefficient of foreign presence (FP) indicates that 1% increase in foreign presence, ceteris paribus, will lead to 4.86E-09 percentage increase in domestic firm productivity.
The interaction of the mean of foreign presence variable with means of technology gap, capital intensity gap and skill intensity gaps in industry level spillover analysis was incorporated in to the model in order to test the effect of foreign presence depending upon the moderating effect of industry absorption capacity. As presented in Table.4.6, the result for the interaction between foreign presence and technology gap shows a positive and statistically significant coefficient at 1% level. This shows FDI presence benefits depend on the industry absorptive capacity.
Figuratively it can be expressed as a 1% increase in foreign presence in industry results 8.39E-10 % increase in industry productivity, in moderate level of technology gaps (i.e., both variables are centered on the means), keeping other factors constant. However, its significance seems low in contributing to domestic firm's productivity as the figure in result shows when it compared to the empirical findings of Blomström and Sjöholm (1999) for Malaysia and Liu et al. (2001) for China.
In fact, this may not be too surprising given that the level of economic development of China and Malaysia is higher, and total FDI in both countries are much higher as compared to Ethiopia.
Even though, it is constrained with the technology gap, it is also similar with the finding of study conducted by Haskel et al. (2002) in UK during the period 1973 to 1992.They estimated a production function for domestic plants augmented with indicators measuring foreign presence in the industry. The estimate suggested that 10 percent increase in foreign presence in UK industry raised the total factor productivity of that industry's firm by 0.5 percent. However, with respect to developing countries cases, the finding is similar with the recent study of Boly et al. (2015) conducted in selected Sub-Saharan countries. They argued on their finding that the foreign presence has positive effect on the domestic firm productivity for those firms who have in better absorptive capacity.
Finally, the result is plausible with the theory that firms' absorptive capacity has shown to be a determinant for FDI productivity spillover occurrence. This explains that the benefit of FDI depends on the industry absorptive capacity and not all industries should be expected to benefit equally from foreign firm presence spillover. There are literatures that support the idea of better absorptive capacity to gain the better spillover effect benefits from foreign presence (Castellani and Zanfei, 2003;Girma and Gorg, 2005;Hale and Long, 2011). To sum up, foreign presence in manufacturing industry of Ethiopia appeared to be determinant of domestic firm productivity, putting absorptive capacity as condition.
The system GMM estimator output in Table 4.6 below shows, horizontal spillover effect has positive expected sign and statistically significant at less than one 1% conventional level of significance. Apart from this, the coefficient of technological and capital intensity gap between domestic and foreign owned firms has negative coefficient, which are similar with the expected signs and statistically significant at one 1% and10% level of significance, respectively. The interaction term variable is also highly statistically significant at less than 1% level of significance.
This implies that the interaction between technology gap and horizontal spillover are positively significant at average level of variables.
This significant positive sign for the interaction term (Horz*TG) indicates that the difference in technology gap between domestic and foreign firms had an important determinant factor for the horizontal productivity spillover occurrence in domestic firms. In other word, this means the effect of horizontal productivity spillover on domestic firm's productivity (VAD) depends on the level technology gap (TG). It implies also that the main effect of horizontal productivity spillover would be the effect of horizontal spillover on domestic firms who had zero technological gaps with foreign firms. Literally it means, a one percent increase in horizontal productivity spillover causes a 0.0107 percent increase in domestic firm productivity for a firm who has technology gaps equals with average technology gap (i.e., has a score of zero on the centered technology gap variable), keeping other factors constant.
This result is similar with the former findings of Kokko (1994), Kokko et al. (1996) and Flores et al. (2000). They argued in their study that certain gap between technology of foreign and domestic firms are necessary for spillovers to occur, for instance, when local firms get an opportunity to copy foreign production techniques or benefit from the training if they have not at least some capabilities to do so, it is impossible to get the benefit as expected. They found a positive result only for domestic firms who have in medium level of technology gap with respect to foreign firms.
Literally, this means, the domestic firms with medium and above absorptive capacity to receive spillover have a benefit and the reverse woks for the firms with lower absorptive capacity. Meyer (2004) also point out that the greater the size of technology gap between domestic and foreign firms, less beneficial FDI is for the host country. It is also similar with the recent finding of Mugendi (2014) in Kenya. He found that skills and technology gap as a major determining factor for horizontal and vertical spillover effect occurrence.
Nonetheless, this finding is the opposite with previous finding of study employed by Turi (2015) regarding horizontal spillover in Ethiopia in manufacturing sector. The reason might be the estimation method difference and the effects of the determinants for horizontal productivity spillover occurrence were not controlled well in his study. The other possible reason is might be the inclusion of foreign firms in his study for analysis. It is also contradicting with the findings of Konings (2001), Aiteken, andHarrisson (1999), who repeatedly come up with no existence of FDI productivity spillover on domestic firms. To sum up, the domestic firm who has technological gap in average level has benefited better from horizontal spillover effects.
As shown in Table 4.6 column 2, the coefficient of capital intensity difference is negative and statistically significant at 10% level, which suggests that a one-unit increase of capital intensity difference from the average level of capital intensity difference results 0.005% decrements in the domestic firm's productivity. The statistical insignificance of the interaction between capital intensity gap and horizontal spillover effect variable, may indicate that the capital intensity gap observed between domestic and the foreign firms is not as such determinant factor for horizontal spillover occurrence in the specified period of study.
The skill intensity gap between foreign and domestic firms, which is proxied by difference between the wage of a worker in a domestic firm and the average wage payment of a worker in foreign owned firms in the same industry, is not statistically significant. Similarly, the impact of industry concentration, which is proxied by the Herfindahl index, are not statistically significant as indicated in the Table 4.6.
As shown in Table 4.6 column 4, except slight changes in the coefficients of estimated parameters, the industry level data gives similar results with firm level data. The forward and backward productivity spillover effects are the interested variables for analysis in this estimation. Since, both of these variables are dummy variables, the domestic firms who did not benefited from forward and backward linkage from foreign firms were used as base or reference category. The result report in tables 4.6 column 4, shows that the backward spillover effect has positive signs as expected and statistically significant at 5% level of significance. This coefficient indicates that industries under which the firms have backward linkage with foreign firms are more likely productive than the industries under which firms had no linkage with foreign firms. To put in more detail, the movement of backward linkage in industry from zero to one percent point produces a 0.00425 percent point change in domestic firms under that industry, holding others factors constant. In contrary, forward spillover, is statistically insignificant at any level of conventional significance level. This indicates that the benefits from forward linkage had not exploited during the period covered by this study These findings are consistent with the finding of Turi (2015) for Ethiopia. It is also similar with the previous work finding of Tomohara and Yokota (2006) in Thailand using firm level unbalanced panel data. They suggested in their study that, on average, FDI improves domestic firms' productivity in the same sector as well as in upstream sectors, but does not affect productivity of domestic firms in downstream sector.
The p-value in Table 4.6 column 4 for the variable labor competition shows statistically insignificant results in any conventional level of significance. As indicated in the descriptive statistics part of this study, the reason for insignificance of labor competition might be the nearly equal mean of the wage rate in both foreign and domestic firms. From this we can infer that the lab our market competition for domestic firms has not resulted because of the presence of foreign firms in the industry.
Finally, on average, FDI improves domestic firms' productivity through the horizontal and backward channels, but does not affect the increase in productivity of domestic firms through forward linkage. The horizontal productivity spillover effect comes to existence through demonstration and competition effects whereas the backward productivity spillover effect realized through local supplier firms related technology transfer.

SYS-GMM Estimation Result Discussion on Determinants of Horizontal Spillover
As shown in the SYS-GMM estimation output table 4.6 column 6 below, the variable such as demonstration and competition effect have positive sign coefficients as expected and they are statistically significant at 5% level of significance. Likewise, the lag of horizontal spillover and technology gap has positive and negative signed coefficients as expected respectively and they are statistically significant at 1% level. The other remaining variables such as skill and labor mobility from foreign firms to domestic firms are not statistically significant in any conventional level of significance. The possible reasons for the insignificance of skill variable are the proxy variable (average wage of workers). As indicated in the descriptive statistics summary, the mean wage rate of workers in foreign firms was less than that of domestic firms, even though; the productivity of workers in foreign firm is greater than workers in domestic firm. From the positive and significant coefficient of demonstration effect, we can suggest that the domestic firms with demonstration effect have 0.84 percentage points more horizontal productivity spillover than firms with no demonstration effect on average, ceteris paribus. Similarly, the positive and significant coefficient of competition effect enables us to suggest that the domestic firms with positive competition effect have 0.67 percentage points more horizontal productivity spillover than firms with no competition effect in average, ceteris paribus. In other equivalent expression, it can be expressed as, domestic firms who used competition positively has 0.67 percentage points more horizontal spillover than firms who fail to use competition positively, or did not face competition.
The negative and significant coefficient of technology gap informed us that an increase in one unit of technology gap result 8.98E-09 units increase in horizontal spillover in average, ceteris paribus.
This implied that domestic firms with low technology gap were bigger recipients of horizontal spillover. Even though, the variable skill and labor mobility was not significant for the possible specified reason above, if workers move from foreign firms to domestic firms, there are possibility to transfer more spillover for domestic firms who had small technology gap. This result is also consistent with the recent finding of Mugendi (2014) in Kenya. He found that technology gap as a major determining factor for horizontal and vertical spillover effect occurrence Source: own computation from CSA survey data

Conclusion
The agenda of attracting FDI in developing countries in recent world has been attracting the attention of many policy makers. Despite the fact that, numerous studies have been employed regarding FDI productivity spillover effects in various countries; the empirical works in Africa is still inadequate (Boly et al., 2015). In this ground and the gaps in former research, the study was aimed to evaluate the productivity spillover effect from foreign owned firms to domestically owned firms in Ethiopian manufacturing sector.
Accordingly, empirical result of this study shows the coexistence of both negative and positive effect arising from foreign presence on Ethiopian manufacturing firms. This implies that foreign presence leads to positive effects in low technological gap firms and negative effect in high technological gap firms with respect to foreign firms. It confirms the theory proposed by Buckley, Clegg and Wang (2010, p.192) for such complicated results of spillover. They argued that complexity of spillover effect challenges the fallacious expectation of equal spillover effect in all firms. The results are consistent with the Boly et al. (2015), Campos and Bruhn (2014), and Bruhn and Calegario (2014) findings that FDI has conditional effect on the host firm characteristics.
Accordingly, we conclude that Moderate technology gap between foreign owned firms and domestically owned firms are crucial determinant factor for productivity spillover occurrence. It is also similar with the result of Girma and Gorg (2005) argument that absorptive capacity is a crucial factor to diffuse the advanced technology to the local industry.
Similarly, this study also shows evidence on connection between the horizontal spillover from FDI and the difference in technology gap. However, technology gaps remain obstacles to FDI spillover for some firms. Local firms with technology gap in average level has benefited from FDI productivity spillover. This study also finds evidences on channels through which the horizontal spillover effect occurs to domestic firm. It occurs through demonstration and competition effects.
In other word, this can explain us that both demonstration effect and competition effect are the determinants for the existence of horizontal spillover effect. The result also indicates that FDI has positive and significant role in an improvement of domestic firm productivity by generating backward linkage in the Ethiopian manufacturing sector.
Finally, we can comprehend from the finding of this study that productivity spillover from FDI to domestic firms in manufacturing firms of Ethiopia is dependent on the level of technological gap.
The competition and demonstration effects and supply related technology transfer are the channels through which horizontal and vertical spillover effects are occurring, respectively. These empirical findings have significant role for policy makers to design strategy which enables to minimize the technology gap between the foreign and domestic firms to maximize the positive impact of FDI in domestic firms via productivity spillover effect. The result shows that the spillover effects in all firms are not equal, so that the policy maker has recommended to give priority for the more promising firms in FDI attracting framework.