 Thank you. Thank you so much This paper is about road infrastructure and enterprise development in Ethiopia And it is co-authored with Mount Soderbaum and Iroh Salim Siva who are just in the other conference room presenting papers and chairing a session and Get in at Alamu from Addis Ababa University And we would like also to acknowledge financial support for this project by the International Growth Center so We have already heard from yesterday's keynote speech that Infrastructure is of course one of the critical Constraints for economic growth in low-income countries. So these countries have Very low infrastructure capital The quality of service is typically low and of course it's very expensive and you have this combination of things Which make it difficult for businesses What's really interesting is the fact that there is some heterogeneity in the impact of infrastructure when we look across sectors So in fact is manufacturing which seems to suffer the most from poor infrastructure in these countries and this is because of the Transactions intensive nature of manufacturing because it has more Backward and forward linkages per unit of value added as compared to other economic activities There are some other authors who also try to relate the firm size distribution of manufacturing with infrastructure So the market for manufactured goods is very low in aggregate terms in developing countries or especially in low-income countries, but it's also highly fragmented into very localized markets because of poor infrastructure so the The market size the firm size in this condition would be typically very low because they are kind of targeting Very small localized markets and when we look at this Description it actually fits very well what we see in sub-Saharan Africa in the sense that the manufacturing sector is still At an incipient stage and it is typically dominated by Small- and micro enterprises now we can turn this argument around and say that if we increase investment in Infrastructure we should be able to witness some strong response from the manufacturing sector because it is a one which has been suffering from poor infrastructure services and Again from yesterday's keynote speech. We have seen this increasing tendency in African countries to invest now in Infrastructure in a very unprecedented way But when we look at the literature there is actually no Strong evidence which kind of investigates or the study which investigates the relationship between you know the distribution and performance of manufacturing firms with Infrastructure services most of the evidence that we have actually comes from developed countries and there is a very excellent review Paper by Aroso and his colleagues in 2010 summarizing this this literature There are few studies coming from developing countries but almost all of them are kind of addressing issues in emerging and semi-industrialized countries like you know Brazil China India and semi-station countries, so We don't really have you know systematic empirical studies on the relationship between infrastructure and manufacturing performance In the case of sub-Saharan Africa there are a few studies of course in Africa, but they seem to be focusing again on The performance and outcomes for rural households like Stefan Birken on Ethiopia and Renko and his colleagues on Kenyan farmers So this is a case study on or an empirical study using Ethiopia as a case country and It's we're kind of excited about this because Ethiopia The situation of infrastructure in Ethiopia is typical of that what we see in sub-Saharan Africa The country has become landlocked since the cessation of Eritrea from Ethiopia in 1993 so all Ethiopian trade now goes through the smaller port of Djibouti and For all practical purposes Ethiopia doesn't have any railway systems or water transport systems So it is heavily dependent on road infrastructure both for public and freight transports And until very recently the road networks were very very poor and The good thing is the Ethiopian government has made road infrastructure one of its top priorities in the last 15 years And the government has implemented three road sector development programs during the period 1997 to 2010 and The total cost of this program has been about seven billion dollars Which was partly funded by international donors, but most of it again coming from domestic revenue sources and This investment has already started to make a big difference both in the quantity and quality of infrastructure in Ethiopia This is the latest report from the Ethiopian roads authority Showing that for instance the proportion of asphalt roads Which are in good and serviceable condition has increased from less than 20 percent to more about 75 percent in 2011 And the quantity of roads in terms of you know road density per thousand kilometers has increased from 24 kilometers to about 50 kilometers So it's significant increase as I said both in terms of quantity and quality of infrastructure So the basic question we ask in this paper is what has been the response of manufacturing firms to this massive investment in road infrastructure by the Ethiopian government and we have two specific questions the first one asking You know, what is the impact on the distribution of firms across different locations? So does you know better road networks make towns more attractive for manufacturing firms And the second question is what is the impact on the startup size of manufacturing firms? We know that the industrial landscape in sub-Saharan Africa is heavily dominated by small and micro enterprises Now we want to see what is the effect on the startup size because the keynote speaker Today has already indicated that if you want really to succeed in manufacturing You need like mid-sized firms to to be active in the manufacturing sector. So those are the two questions We would like to address This is a preliminary evidence showing, you know, the effect of well not the effect But that's that some evidence showing the change in the distribution of Manufacturing firms in Ethiopia So the blue line for instance shows you the decline in the share of the capital city Addis Ababa in terms of you know Manufacturing firms does decline from about 65 percent in 1996 to just about 40 percent in 2009 and the share of the top five towns in general has also dropped very significantly what this means is essentially That towns which have never been very important for manufacturing now are kind of attracting manufacturing firms And the concentration of manufacturing is kind of declining over time so We have a very Interesting and very exciting data sets which we can brag about We're basically combining two data sources one of them is a GIS based town level panel data on Road accessibility. So this is using geographic information system and it is as I said Sound level indicator and we also have firm level panel data on Ethiopian Manufacturing coming from our collected by the Ethiopian statistical agency This GIS data is very interesting because it starts off from geo-coded project level information Provided by the Ethiopian roads authority So we have information not only in terms of you know the quantity of roads But we also have you know the pavement type and the condition of the roads themselves. So And we use two GIS analytical tools the so-called you know accessibility analysis which Has two parts in it the service coverage analysis and the origin destination analysis So instead of just simply reporting how much government has spent on road infrastructure or what is the quantity of roads? We're kind of looking into what is the impact on you know road accessibility. So it's more of a quality indicator if you like than simple Crude indicators that has been used in previous studies. So based on this GIS tools we have three indicators of road accessibility The first one is area accessible as you see which is simply the total land area that can be accessed by road During a one-hour travel from the towns from the center of the town So we just basically looking into a five kilometer buffer zone on both sides of the road and looking into you know How much? area you can access In a one-hour travel so that is one of our indicators the other one is travel distance again looking into how far you can Travel in one hour again from the center of the city Using all the roads that serve a particular town. So that is the second indicator The third one is what we call the travel time to major economic destination. So we have identified about 15 economic centers 10 of which are capital cities of regional states and we have a few commercial centers So we're basically saying you know what the physical distance of course to this major economic destinations is a fixed Factor it doesn't change but the travel time could change as a result of road accessibility. So those are the three Indicators we have and this is the time trend of these three variables Which would be easier if I show you in this graphical form The green line at the top shows you the decline in travel time to major economic destinations It has been declining by about five hours per annum The red line tells you the travel distance so you can now travel like you know 46 kilometers more in 2009 as compared to 1996 and The blue line shows you the area accessible which again has increased significantly over the study period So this is for instance one One of the GIS tools which allows us to calculate the area accessible. So it's just using you know buffered on Analysis which is which is really exciting The firm level data as I already indicated comes from The annual census of manufacturing carried out by the Ethiopian statistical agency It covers all manufacturing firms that employ at least 10 workers So it basically looking at formal sector manufacturing firms So that is that's our database for manufacturing firms and It has also geographic indicators So we know in which town and in which region each firm is located So we kind of combine these two data sources using these geographic indicators This is our basic Econometric model on the left hand side is the logarithm of the number of manufacturing firms the total number of manufacturing firms I Identifies the town T identifies time Rn is our indicator for road networks and x is a bunch of control variables Now if you apply ordinary list squares to this model the coefficients are of course going to be biased and inconsistent This is simply because The information which is used by government to assign red road projects has a lot of overlap With the information used by manufacturing firms to choose their location So it would be very difficult to identify Whether firms are locating in that particular area because of you know roads or is it because you know economic potential Which the government also uses to assign road projects So we try different estimation methods to overcome this endogeneity of road placement The first one is using the fixed effects estimator Which would be excellent if these road assignment criteria by government are time invariant fixed effects Okay, so you just difference that out and that would be One way to resolve the endogeneity problem But the problem with this With an estimator is that it doesn't capture the agglomeration the dynamic agglomeration effect, which is quite important for manufacturing firms So what we did was we used the Blunderland bond system gmm estimator Which kind of includes the lagged number of firms on the right hand side of the equation And also uses instrumental variables to resolve the endogeneity problem So we can discuss the technicalities of this This model the other approach we used is the so-called proxy variable approach. So if We are we convince ourselves that you know road assignment by the Ethiopian road authority is based on Observable characteristics Then we can include those characteristics directly into the econometric model So x in this model could be expanded to include the the criteria with the Ethiopian government uses to allocate roads So that's one way by which you can you can resolve this So when we ask the Ethiopian road authority on what criteria they use to assign roads The first thing is we didn't find any publicly available information So we really have to ask, you know the managers to tease out What exactly they do and most of the time they they told us they assigned 20 percent for economic development potential another 20 percent for Roads that go to food surplus or cash production Areas basically looking into agricultural potential and connectivity of roads So the problem with this Criteria is we don't really know how exactly they operationalize these indicators. So how do you measure economic potential? How exactly do you capture food surplus? Production so we have to come up with our own proxy for The criteria which the road authority told us Are used to assign roads the first one is agricultural potential We identify regions or districts which have been served or included in the public safety nets program This is a joint a donor funded Program, which is also led by the Ethiopian government. I should say To provide support for regions which are suffering from chronic food shortages So if you are part of this psmb You would be of course located in a low agricultural potential area If you are not included in this program then you must be either food self-sufficient or food self surplus producer And then we also try to control for initial conditions by including the average number of firms during the period 1996 to 1998 And then we have of course population which is also indicated as another as one of the indicators and regional fixed effects Now to the results. So I'll first show you a simple ordinary list of squares estimates Of the previous model, but averaging the total number of firms during the period 1999 to 2009 And okay, and and also for the average for the Road accessibility indicators as well So this is the result The first column shows you The coefficient on road accessibility. That is the logarithm of SEC means area accessible during the period 1999 to 2009 The coefficient is very high if you don't include any control variables Once we include Approxies for road assignment criteria the coefficient declines significantly to about 0.47 This essentially shows you that you really have to be very careful about Indogenous assignment of road projects. So that's one of the lessons we learn And of course the coefficient of lan n 96 99 that is a number of the initial number of manufacturing firms. It's also very significant and Positive showing you that there is of course some persistence in the number of manufacturing firms This is again the same oil less model using travel distance as Indicator then again, we see some positive and significant impact Now let's subject our model to a more strict estimation method. That is a panel fixed effects So you see that the coefficient right here On sec lan sec is positive and significant the same with travel distance at 0.348 And tt or d travel time to from origin to destination has a negative coefficient Suggesting that you know reducing travel time actually leads to an increase in the number of manufacturing firms That's exactly what we expect to see From this kind of analysis So we're very confident that this thing is capturing a very interesting relationship between infrastructure and number of firms And then we subject our model again using the system gmm estimator again, what you see is travel distance And area accessible have a positive and significant impact on the number of manufacturing firms And travel distance as negative, but the coefficient has become less precise in this in this in this specification And the lagged dependent variable This is the lagged number of manufacturing firms again has positive and significant the coefficients telling you again The agglomeration effects But of course the coefficient is less than one which means Places which have never been important for manufacturing before are now Experiencing a faster increase in the number of firms Now the previous three results basically tell you the net change in the number of firms Which is of course the result of entry and exit So what we want to do now is focus entirely on the total number of entrants. So that's gross entry and This is usually a very small number usually the average number of entrants is about three and Most towns actually have or a large proportion of towns have zero number of entrants So usually you would use what we call count data models the poisson model But you have to kind of take into account The fact that most towns have zero entrance. So we used what we call zero inflated negative binomial model And the results shows again a very positive and statistically significant effect of road accessibility Both in terms of you know area accessible and travel distance But we don't see any significant impact on the total number of firms as a result of you know reducing travel time to major economic destinations The last Specification or the last research question we have is the effect of Road infrastructure on startup size. So the dependent variable in this case Is the size in terms of number of workers of a startup firm K in town i at time t as a function of road networks and a bunch of control variables and The results are what you see here A positive and significant impact when we look into area accessible and travel distance And of course reducing Travel time to major economic destinations also increases the startup size of manufacturing firms, which is really really very interesting And these two variables right here in terms of you know, the logarithm of district level population also has a positive and significant impact showing you that demand plays A very important role This specification includes the initial number of firms as one of the control variables And you can see that Although the initial number of firms Has a positive and significant association with the total number of firms The startup size is actually lower If you locate in You know in dense or highly intensive activities Towns with intensive manufacturing activities. So it's an indication of the intensity Of competition if you locate yourself in major centers of manufacturing And then we used the two-stage list of squares Estimator i'm done. Thank you This is you know just to Take into account the indogeneity of road placement again Our instrument in this case is road density in 1990 As instrument for road networks after 1999 So this should affect firm entry The size of entrance Through road networks, but not directly. Okay, so this is because it's an individual Entrepreneur's decision. So it doesn't have to do with The indogeneity issue So we have seen that the road sector development program has made a significant improvement In the quality and quantity of road networks in Ethiopia And of course the number of manufacturing firms has increased during the sample period But we see significant difference across towns In the sense that towns which have never been important for manufacturing purposes have now Become attractive for investors because they are now well connected With the country's road networks now historical centers of manufacturing still are very competitive and attractive for manufacturing firms So as a result of what we have seen the concentration the geographic concentration of manufacturing has has started to decline So this is very important and of course the startup size Is higher in For firms which locate in towns with better road networks I'll leave it there and thank you. Thank you. Thank you very much. Um, Admasi. Yeah