Different Obstacles for Different Productivity Levels? An Analysis of Caribbean Firms

This chapter characterizes several aspects of firms in the Caribbean using two sources of microdata: the World Bank Enterprise Survey and the Productivity, Technology, and Innovation in the Caribbean. These data offer a fresh opportunity to investigate and test theories about productivity and economic growth in the Caribbean. The chapter first presents a comprehensive portrait of firms in 14 Caribbean countries from several angles: size, sector, age, interconnectedness, export status, and human capital. These attributes are then related to the productivity levels of the firms and the obstacles they face in their current operations. In some Caribbean countries firms of differing productivity levels perceive varying obstacles as the most constraining to firm performance; whereas in other countries, firms report the same dominant obstacle regardless of productivity level. Since these obstacles could be constraining productivity growth, the chapter offers a method for analyzing the most binding constraint identified by firms in different segments of the productivity distribution.

Much of the literature relating fi rm characteristics to productivity and growth in Latin America and the Caribbean (LAC) either lumps Caribbean countries into one observation or tends to overlook them altogether. This is not because researchers want to exclude the Caribbean, but because the data defi cit that often poses a challenge for the LAC region is even more extreme when it comes to Caribbean countries. Only a small fraction of over 100 identifi ed indicators affecting growth are available for these countries. Further, limited availability of household data or fewer observations on fi rms is often prohibitive for standard methodological analysis of economic growth .
So, is an independent analysis of Caribbean fi rms even needed? The simple answer is yes. Small population size, geographical characteristics, and main economic activities set Caribbean economies apart from most Latin American economies. As with Latin America, there is stark heterogeneity among and within Caribbean countries. Even though size may be a defi ning factor, it is lower productivity levels that defi ne the declining growth of Caribbean economies relative to other small-sized economies . Increasingly, understanding macroeconomic trends requires an understanding of fi rm dynamics at a micro-level and productivity levels within and across industries (Syverson 2011 ). Therefore, Caribbean policymakers need this type of micro-data, analysis, and dissemination of information tailored to the region at their disposal. Recent fi rm-level data from the World Bank Enterprise Surveys (WBES) and the Productivity, Technology, and Innovation in the Caribbean (PROTEQin) Survey offer new opportunities to understand better the characteristics of Caribbean fi rms at different levels of productivity and the challenges or obstacles that they face in their daily operations.
The primary objective of this chapter is to fi ll a void in the literature about fi rms in the Caribbean through a comprehensive analysis of different fi rm characteristics and productivity. These fi rm-level characteristics are discussed at length in this book, but they are not directly applied to the Caribbean context. 1 In the next section, we briefl y contextualize the Caribbean economies within which fi rms are operating. Then we examine some of the basic fi rm characteristics that are frequently empirically linked with productivity, such as fi rm size, sector, age, exporter status, and use of information and communication technologies (ICT). We then deepen the analysis by specifi cally focusing on human capital, looking at both management and employees. The following section shows our analysis of fi rm-level labor productivity in relation to the main characteristics of fi rms (fi rm size, sector, age, exporter status, and ICT usage). Then we investigate the obstacles reported by the fi rms surveyed, specifi cally looking at the correlation and variation between obstacles and fi rm performance. Most of the analysis draws on the most recent wave of the WBES, which was carried out for the fi rst time in 14 Caribbean countries in 2010. 2 We complement the analysis with data from the PROTEQin, which was conducted for the fi rst time in 2013 in fi ve Caribbean countries (Barbados, Belize, Jamaica, Guyana, and Suriname).
Overall, the results from the Caribbean micro-data tell a familiar story about fi rms with lower productivity levels-they tend to be smaller, to export less, and to have less human capital and technological inputs-but they also tend to report different obstacles to their current operations. If private-sector-led growth is expected to bolster the economy, then documentation and dissemination of the characteristics of this sector and the DIFFERENT OBSTACLES FOR DIFFERENT PRODUCTIVITY LEVELS? … 209 bottlenecks that lower and higher productivity fi rms are facing seems to be a necessary fi rst condition for apt policymaking.

LANDSCAPE OF THE CARIBBEAN ECONOMIES
This chapter uses data from 14 Caribbean economies. Table 7.1 shows that all of the Caribbean economies discussed in this chapter meet the defi nition of a small economy 3 (except the Dominican Republic) based on having a population of less than three million people. The majority of the economies have a population of less than one million people. The econ-  $14,132 (in PPP terms). The majority of the countries are islands where tourism is the largest industry. The number of ports is included in Table 7.1 to demonstrate the interconnectedness of the region to world trade networks, with the vast majority 4 of the trade being transported by sea (Kaluza et al. 2010 ). Naturally, these economies face a small domestic market and can be at a disadvantage in global markets, although size does not have to be a binding constraint. Low productivity levels in the private sector in the region compared with similar small economies is a pressing concern for the future of the Caribbean  ).

Box 7.1. Recent developments in data collection in the Caribbean
The release of the 2010 WBES was a starting point for comparable fi rmlevel data in the Caribbean. However, from the outset, researchers recognized the need for subsequent surveys in order to analyze the evolution of fi rms in the region. Fortunately, not too long after the fi rst WBES was conducted, the region implemented the fi rst wave of pseudo-follow-up surveys-the PROTEQin. This survey was commissioned by the Inter-American Development Bank (IDB), with funding from the Compete Caribbean Program, a regional private sector development and technical assistance initiative fi nanced by the IDB; the United Kingdom Department for International Development; and Canada's Department of Foreign Affairs, and Trade and Development. The survey was executed in partnership with the Caribbean Development Bank. 5 Administered between 2013 and 2014, the PROTEQin is a critical development in terms of data collection in the Caribbean and targeted establishments that were covered by the 2010 WBES in fi ve economies: Barbados, Belize, Jamaica, Guyana, and Suriname. This decision allowed researchers a fi rst opportunity to use panel data in analyzing fi rm-related issues in the Caribbean. 6 DIFFERENT OBSTACLES FOR DIFFERENT PRODUCTIVITY LEVELS? … 211

PRINCIPAL CHARACTERISTICS OF THE FIRMS IN THE CARIBBEAN
The dominating characteristics of the fi rms surveyed in the Caribbean are that they tend to be micro or small, concentrated in the services sectors, mature, and non-exporters. The documentation of the proportions of WBES fi rms with these attributes in each of the countries illustrates the heterogeneity between Caribbean countries and serves as a starting point for the rest of the chapter, establishing the particular features of the fi rms that are often linked to productivity in the literature. The majority of fi rms are small (11 to 50 employees) or micro (10 employees or less). Figure 7.1 shows that 54 % of the fi rms in St. Vincent & the Grenadines are micro and 38 % are small for a total of 92 %; a little over 6 % of the fi rms are medium and just a shade over 1 % are large. Very few large fi rms exist in any of the Caribbean countries. The Dominican Republic is the only country where more than 10 % of the fi rms in the WBES are large enterprises. As we expect, typically countries with very small populations have a relatively greater percentage of micro and small fi rms. These countries appear on the left side of the graph, but there are some exceptions. For example, Barbados has a smaller population than Belize but a greater proportion of medium fi rms.
In a recently published note that maps the enterprises in LAC based on WBES data, there tend to be even more small and medium enterprises (SMEs) in the Caribbean than in the rest of LAC (94 % versus 90 %) and more fi rms are in the services sector  ). 7 Although favorable views of SMEs contend that they spur competition and are a good source of employment, this argument only holds if the SMEs are productive, which implies that they are competitive and innovative (Pagés 2010 ).
The PROTEQin expands the scope of the WBES while also incorporating more detailed questions related to labor, productivity, technology and innovation for 727 fi rms. The dataset provides updated information on how fi rm characteristics and performance have evolved since the 2010 WBES. For this reason, we intersperse fi ndings from the PROTEQin where possible to provide more recent information for selected countries and to check the robustness of the WBES data.
In fact, research using the global WBES found that, while small fi rms may have the largest shares of job creation and sales growth, large fi rms tend to display higher productivity growth (Ayyagari et al. 2011 ). Chapter 3 showed that large fi rms are more likely to invest in innovation and that those that do are more productive.
In most countries in the Caribbean, there is a greater proportion of fi rms in the services sector; however in Suriname, for example, fi rms are split roughly evenly between the services and manufacturing sectors. In the WBES, the fi rms self-classify as either being in manufacturing or in services. The corresponding workforce within the countries may be even more heavily concentrated in the services sector. In the LAC region, over 60 % of the workforce is in services; in the Organization of East Caribbean States, the number is over 80 % (Caribbean Knowledge Series 2013 ).
As discussed in Chap. 1 , and in line with recent research on productivity growth, it is the services sector that drags down overall productivity levels in LAC (Pagés 2010 ). Several studies have looked at the differences in productivity and innovation in the two sectors (Arias Ortiz et al. 2014 ;Crespi et al. 2014 , for Latin America only; Arias Ortiz et al. 2012 ;IDB 2011a ). These studies found that the allocative effi ciency in the services sector tends to be much lower than in manufacturing. Knowing the sectoral composition of the fi rms in each country is a key element for analyzing the productivity of fi rms. In addition to being small and largely in the services sector, Caribbean fi rms also tend to be older. Very few new fi rms (defi ned as less than three years old) exist, whereas mature fi rms (defi ned as those in existence for over ten years) are much more prevalent. Mature fi rms represent the majority, except in Dominica, where the proportion of such fi rms dips below 50 %. In general, the LAC region tends to have a smaller proportion of young fi rms compared with other developing regions . The implications of age and productivity could go in either direction. Young fi rms are often seen as being a potential source of newness and innovation; however, mature fi rms may be seen as having stood the test of time.
The next important question relates to how connected these fi rms are. Given that fi rms tend to be smaller and older, have they adopted ICTs to connect to domestic or international markets? Are they internationally engaged? Figure 7.2 shows that cellphones and email are widely used in everyday business practices. More sophisticated ICTs, such as owning a website, which often requires some basic programming knowledge, are much less pervasive. There is a lot of heterogeneity within the Caribbean with regard to ICT, as there is throughout LAC, where evidence suggests that within-country differences are as notable as between-country differences (IDB 2011b ). The low levels of website ownership by tourism-based economies suggest that there is room for gains by attracting new clients who are not being reached by conventional hotel and restaurant search methods. The PROTEQin provides updated information about ICT penetration in select Caribbean countries. Firms were asked the same series of questions about email, websites, and cellphones for business operations. The improvements in these indicators vary by country. Countries like Barbados and Belize, which already showed relatively high levels in 2010, saw slight improvements in websites (Barbados and Belize) and cellphones (Barbados). With a 4 % improvement over 2010 in cellphone usage, Barbados reached 100 % penetration in both cellphone and email usage to communicate with clients. Suriname and Jamaica showed signifi cant improvements in ICT usage between the two survey periods. For example, in Jamaica, cellphone use increased by 24 %. Suriname saw sizeable increases in both website usage (28 %) and email usage (18 %). Guyana is the only country that showed declines in ICT penetration in both website and email usage. On the whole, for cellphone penetration, the PROTEQin shows improvement over the WBES 2010 average, with all fi ve countries above 90 %. 8 To meet regional averages, Jamaica, Guyana, and Suriname need to improve email usage. Use of fi rm websites was by far the weakest area for the selected countries, with Barbados being the only one to outperform the 2010 regional average. Despite these gaps, the large improvements between 2010 and 2013 in some of the underperforming countries, like Jamaica and Suriname, should not be overlooked.
In terms of international linkages, the WBES data shows that most Caribbean fi rms are non-exporters ( Fig. 7.3a ); therefore, a very small proportion of sales are derived from either indirect or direct exports ( Fig. 7.3b ). In general, exporting is thought to be positively linked with productivity. A recent survey of micro-econometric studies from 33 developed and developing countries summarizes corroborative evidence from 1995 to 2004 (Wagner 2005 ). The author claimed that most of the differences were due to pre-entry self-selection into export markets rather than gains in productivity post-entry into the market (Wagner 2005 ). 9 In addition to whether or not a fi rm is exporting, the average proportion of sales earned from exports ranges from 3 % in Grenada to 16 % in Dominica. Across the Caribbean, a very small average proportion of sales are being generated from indirect export sales.
Given the dominating characteristics of the fi rms covered in this section (small, old, and in the services sector), if policymakers want to help fi rms become more internationally engaged and connected through technology, DIFFERENT OBSTACLES FOR DIFFERENT PRODUCTIVITY LEVELS? … 215 preparatory work along the supply chain is needed on the pre-entry side. For example, in the Caribbean, even fewer fi rms are engaged in indirect export sales than are in direct export sales.

PRINCIPAL CHARACTERISTICS OF HUMAN CAPITAL IN THE CARIBBEAN
Every fi rm is made up of its people. Just as aggregate productivity is the combination of the productivity of individual fi rms, each individual fi rm's productivity is the sum of the productivity of its workers. In this section, we delve into the characteristics of human capital in Caribbean fi rms, from managers down to workers. Unobservable factors such as the skills of the workforce and managerial capability are often more responsible for the variation in fi rm performance than are observable fi rm attributes such as size, age, and international linkages (Jensen and McGuckin 1997 ). A better understanding of the knowledge, capabilities, and background of the workforce is important, as both the observable and unobservable characteristics of a fi rm must be included in a complete analysis of fi rm growth (Laursen et al. 1999 ).

Entrepreneurs in the Caribbean
We begin with an analysis of the entrepreneurs (fi rm owners or managers) in the Caribbean. Recently, some scholars have attributed entrepreneurship with the commercialization of new knowledge and consider it a third driver of economic growth (Vivarelli 2013 ). 10 Although the economic literature has long been fascinated with entrepreneurship, not all characterizations describe entrepreneurs as agents of change and economic growth (Wennekers and Thurik 1999 ). 11 Although new businesses may contribute to job creation, in order to contribute to productivity, businesses must also grow into their potential (Wagner 2014 ). Figure 7.4 shows that fi rms in the Caribbean are not often created to introduce a new product or idea. Coupled with the fact that fi rms tend to be mature, this suggests that the majority are not responsible for commercializing new knowledge that would position them as drivers of economic growth. Figure 7.4a shows whether the fi rm was established out of necessity; the responses vary widely across countries. Figure 7.4b shows what type of opportunity motivated the fi rm's creation. For example, more than 80 % of the fi rms interviewed in Suriname reported that the business was started because of a lack of better employment opportunities. In contrast, none of the fi rms in Dominica responded that this was the case. 12 Figure 7.4b shows that fewer fi rms were created to develop a completely new product or idea than to replicate or modify an existing product or idea. The results are similar for both Caribbean countries and Latin American countries (see Chap. 6 ). These fi ndings are consistent The previous experience of top managers varies widely throughout the Caribbean. In Fig. 7.5 a signifi cant number of countries have some top managers that transitioned from being unemployed into the position. In Suriname, for example, over 10 % of those surveyed transitioned from being unemployed to being employed as a top manager, but this does not seem to be the general trend. On the whole, the top manager tends to have previously held a managerial position that may have provided the impetus to start a new business, especially given that the majority of fi rms are created to either imitate or replicate existing products or services.
In sum, Caribbean fi rms tend to replicate, imitate, or differentiate products or services that exist in the market. Further, very few of the fi rms surveyed are considered high-growth ones. The role of the entrepreneur in transforming an economy rests on the match between available market opportunities and entrepreneurial talent (Naudé 2008 ). 13 So, if most of the entrepreneurs in the Caribbean are not commercializing new knowledge, they may be absorbing technology from elsewhere, which requires social capacity to imitate and differentiate-skills also associated with gains in productivity. These skills relate to a broad variety of factors within economies, including but not limited to the general level of education of the workforce, the technical competence of workers, and the amount of technical training provided to workers.

The Caribbean Workforce
Since 1960, there has been a lot of progress in the Caribbean in terms of attaining primary and secondary education. The region's average years of schooling for the adult population are now on par with the rest of Latin America and approaching Organisation for Economic Co-operation and Development (OECD) averages. The same is not true of transition from secondary to tertiary education. In the fi rms surveyed, the average percentage of workers with at least a bachelor's degree ranges from 2 in Grenada to 20 in the Dominican Republic. 14 In addition, pass rates for math and English tests are often below 50 %. These signs point to a deeper issue of whether there is a match between skills taught in school and those demanded by employers in the workplace (Caribbean Knowledge Series 2013 ) (Fig. 7.6 ).
Finding workers with the right skillset is a major issue in the Caribbean, where over 35 % of fi rm owners report having unfi lled vacancies. An inadequately educated workforce is one of the most often cited obstacles to fi rm growth in the region (see "Obstacles to Firm Operation in the Caribbean" below). The "right" skills, however, differ by country. On the one hand, in Grenada, Barbados, and Antigua and Barbuda, workers with technical skills are more diffi cult to fi nd. On the other hand, in Guyana and the Dominican Republic, employers have a slightly harder time fi nding workers with social skills.
An interesting fi nding from the PROTEQin data is the variation in the diffi culty of fi nding certain skills by job type (i.e. managerial versus professional). For example, the PROTEQin asks fi rm owners to rate the diffi culty of fi nding candidates with appropriate skills by different positions within the fi rm. Figure 7.7 displays the percentage of fi rm owners who responded that certain skills were very diffi cult or almost impossible to fi nd in candidates. 15 The fi ndings are notable. Adequate job-related skills tend to be the most diffi cult attributes to fi nd in candidates for both managerial and professional positions. On average, almost 30 % of fi rm owners in this subsample found core skills to be very diffi cult or almost impossible to fi nd when hiring professionals compared to one-fi fth when hiring managers. These results show that, in the Caribbean, there is a lack of adequate skills not only for lower-level workers, but also when seeking capable managers.
When fi rms were asked in the PROTEQin to identify the importance of various factors causing skill shortages, 52 % cited worker emigration as important, very important, or critical. Considering in the Caribbean net migration is among the highest in the world and that outfl ows are predominantly migrants with a tertiary education (Nurse and Jones 2009 ), 16 it could even be surprising that only 52 % of fi rms cited worker emigration as such an important factor. It is possible that the diaspora has come to be seen in the Caribbean as a unique source of human capital that provides links to external markets and international customer bases, transfers industry-specifi c knowledge, and acts as sources of investment. 17 Evidence from a recent report suggested around 40 % of the diasporic entrepreneurs surveyed, 18 19 indicated that they earned some form of revenue from clients in the diaspora. In addition, interviews with large iconic fi rms in the Caribbean (Suriname, Jamaica, and Guyana) and diasporic fi rms outside the Caribbean (e.g. New York) revealed these large iconic fi rms have designed business strategies to target the diasporic customer base (Nurse and Kirton 2014 ) who then also infl uence consumer taste in the international markets where they have migrated. While the majority of fi rms responding to the PROTEQin acknowledged that emigration may deplete local human capital resources, causing skill shortages, they more frequently cited the quality of education or a shortage in the number of local professionals trained by local institutions., 20 21 They also noted that emigration of workers may provide intangible inputs to local business development, especially through their potential link to an international network and potential customer base outside the country.  Despite managers and skilled workers having relatively high levels of educational attainment in Barbados, Belize, and Jamaica, over 60 % of fi rms in those countries cite a lack of a strong educational background as a major or severe obstacle to productivity. This may be an indication of a mismatch between the skills students are learning in school and the skills desired by the employers in these countries rather than a refl ection of low educational attainment. These workforce constraints are less of an issue in Guyana and Suriname, where only around 30 % of fi rms cited lack of educational background as a major or severe obstacle. This does not mean that it is not an important issue for fi rm productivity, just that there are likely other, more pressing, obstacles in the fi rm manager's mind.
DIFFERENT OBSTACLES FOR DIFFERENT PRODUCTIVITY LEVELS? … 223 below the average. Then, we analyze the relationship between different fi rm characteristics and higher performing fi rms versus lower performing fi rms. We repeat the exercise for the subsamples of fi rms in the manufacturing and services sectors. Table 7.2 shows different patterns for higher and lower productivity levels in relation to various characteristics of fi rms. Relatively higher productivity fi rms tended to be larger at the end of 2009 and to have more employees in 2007. The pattern is the same for manufacturing and services, but the average size of manufacturing fi rms appears to be larger than services. The higher productivity fi rms are also generally older, have a greater proportion of sales from direct exports, and have more than 10 % foreign ownership. In manufacturing, the higher productivity fi rms are older, on average, than services fi rms. In the services sector, higher productivity fi rms have a greater concentration of foreign ownership. In addition, the proportion of higher productivity fi rms with direct exports is greater in manufacturing than in services. Human capital patterns are as we might expect. Managers of relatively higher productivity fi rms tend to have slightly more years of experience, on average. Interestingly, the proportion of full-time permanent employees with at least a bachelor's degree is the highest in the relatively higher productivity fi rms in the services sector.
For technological absorption capacity and usage, different patterns emerge for manufacturing and services based on having a website, using foreign technology, or having an internationally recognized certifi cation. First, a very small proportion of fi rms in the services sector use technology licensed from a foreign company. Second, a greater proportion of fi rms in manufacturing have an internationally recognized quality certifi cation (19 versus 16 % in services). Third, a greater proportion of services fi rms have their own website, which is likely consistent with how critical it is to share information with clients. As mentioned earlier, for fi rms in services, such as hotels and restaurants, online advertising is becoming increasingly important. Across the board, a greater proportion of higher productivity fi rms have a website, use foreign technology, or have an internationally recognized certifi cation.
As a robustness check, we conduct a similar analysis using the PROTEQin data. We fi nd many of the same results using the much smaller subset of data. 22 However, we note a few interesting deviations. First, fi rms in the fi ve countries are generally slightly older than in the Caribbean as a whole. The average age of low-productivity fi rms in this subsample is four years older than the average age of low-productivity fi rms in the 14 Caribbean countries surveyed for the 2010 WBES. The differences are even starker on a sectoral basis, where fi rms that maintain higher productivity levels are, on average, ten years older than low-productivity fi rms in the manufacturing sector (the gap is six years in the services sector). In other words, the most productive manufacturing fi rms tend to be the oldest in this subsample and unproductive services fi rms tend to be the youngest. In terms of exports, the PROTEQin data shows a much wider gap between the percentage of exporting fi rms above and below-average productivity levels (33 % of above-average manufacturing fi rms export versus 18 % of below-average fi rms). Interestingly, the reverse is true for services, with a larger percentage of low-productivity fi rms exporting (11 %); only 8 % of services fi rms with average or aboveaverage productivity levels are exporters. This may signify that services fi rms are beginning to export before they have the required internal capacities, thus hindering their productivity.

OBSTACLES TO FIRM OPERATION IN THE CARIBBEAN
The WBES asks fi rm owners a series of questions about their perceived obstacles to current operations. The goal of these questions is to identify particular aspects of the business environment that are constraining fi rms. Subjective measures of the perceptions of obstacles have been found to be highly correlated with objective measures for the Caribbean specifically (Ruprah and Sierra 2013 ) and worldwide (Hallward-Driemeier and Aterido 2009 ). Therefore, in this chapter, we do not question the validity of a particular obstacle being seen as the biggest one, rather we take the fi rms at their word. Figure 7.8 presents the frequency with which interviewed fi rms in selected Caribbean countries reported each of the 15 identifi ed obstacles in the survey.
Caribbean fi rms identify different obstacles as the biggest, which assumes that the most frequently reported obstacle in a country is the one considered to be the most constraining for the majority of fi rms. For example, in the Bahamas, the highest percentage of fi rms (34 %) reports lack of an educated workforce as their biggest obstacle; whereas, in Barbados the highest percentage (28 %) identify access to fi nance as their biggest obstacle to operations. In Jamaica, 34 % of fi rms identifi ed tax rates as their biggest obstacle. A recent, in-depth analysis by Nugent and Schmid ( 2014 ) confi rmed that Jamaica's tax system has long been an issue in the country.
The 2013 PROTEQin data also enabled us to analyze whether the perceptions of primary obstacles have changed for the fi ve PROTEQin countries since the 2010 WBES. Table 7.3 shows that, for the most part, the main constraints on fi rm growth remain unchanged in Belize and Jamaica. While the same holds true in Suriname, the severity of the obstacle (an inadequately trained workforce) decreased by almost two-thirds from 2010, and fi rms began to cite a wider array of obstacles in 2013. In Barbados, electricity replaced access to fi nance as the biggest obstacle, and in Guyana, electricity replaced an inadequately trained workforce.
Overall, access to fi nance, a trained workforce, electricity, and tax rates were consistently cited by fi rms as primary, secondary, or tertiary obstacles to fi rm growth in 2010 and 2013. Possible explanations for the shifts in the relative importance between particular top obstacles may stem from other changes that the fi rms in the group surveyed have experienced in the three-year period.

Are Obstacles Different for Firms in Different Productivity Quintiles?
The biggest obstacles identifi ed in Figure 7.8 are for all of the fi rms in a given country. Hallward-Driemeier and Aterido ( 2009 ) found that fi rm characteristics had an impact on the relative importance of obstacles and highlighted the need to look at differences within countries based on fi rm characteristics. In this vein, we separate fi rms by labor productivity quintiles to explore whether fi rms with different levels of labor productivity in the Caribbean identify different primary obstacles to operations. First we divide the fi rms evenly into fi ve productivity quintiles where one is the lowest productivity quintile and fi ve is the highest. Table 7.4 visually presents the frequency with which the biggest obstacle is identifi ed by a particular quintile of productivity. In the Bahamas and Barbados,    Source : Author's own elaboration based on WBES data Notes : Productivity is measured by the fi rmʼs sales for the last fi scal year defl ated to 2009 US dollars and divided by the total number of permanent full-time employees plus the total number of temporary full-time employees (adjusting for the number of months). The quintiles are divided evenly within each country, so there are roughly the same number of fi rms in each quintile. Quintile 1 represents the lowest productivity fi rms and Quintile 5 represents the highest productivity fi rms in each country. For each quintile, the number represents the percentage of fi rms declaring the corresponding category as the biggest obstacle the majority of fi rms in the lowest productivity quintile identify access to fi nance as their biggest obstacle. 23 In the Bahamas, as the productivity quintiles increase, the majority of fi rms identify an inadequately educated workforce as their biggest obstacle. It is possible that these shifts in identifi ed obstacles refl ect increases in exposure to a wider gamut of obstacles as a fi rm becomes more productive and is faced with more challenges to growth. In the Bahamas, for example, the medium productivity quintile displays an even split between fi rms that report customs and trade and an inadequately educated workforce as the biggest obstacles. A possible explanation for this could be that these medium productivity fi rms are at the stage when they are just beginning to export. The relatively lower productivity fi rms may not be attempting to export yet and thus are not affected by such regulations, while higher productivity fi rms may already have mastered the red tape and no longer see that as an obstacle. In other countries, such as Jamaica and Trinidad and Tobago, the dominance of a particular obstacle does not differ by productivity quintile. This suggests that the country's business environment may have a feature that affects all businesses and may trump the degree to which the biggest obstacles change according to fi rms' characteristics.

Econometric Model
Within economies and within industries, some fi rms are simply more efficient than others. Using the same measured inputs, high-productivity fi rms (in the 90th percentile) outperform low-productivity fi rms by 2:1.
In India and China, the ratio has been found to be as high as 5:1. Further, within-industry dispersion has been found to be on the rise in the United Kingdom (Syverson 2011 ). Recent publications about innovation and productivity in the LAC region found that fi rm-level productivity is heterogeneous even within specifi c economic sectors (log productivity differences between the 90th and 10th percentile were found to be 2.66 log points in services and 2.53 log points in manufacturing). Theory often attributes this to market frictions that can be exacerbated by weaknesses in the institutional environment (Arias Ortiz et al. 2014 ). The following analysis uses quantile regression techniques following the methodology described in Goedhuys and Sleuwaegen ( 2009 ). This methodology is relevant given that the distribution of the dependent variable, labor productivity, is skewed. 24 , 25 Our interest is in the relationship between perceived obstacles to operation and productivity changes among the different quantiles of the distribution. Since a large number of observations are needed for this type of analysis, we pool the responses of all the fi rms in all 14 Caribbean countries surveyed for the 2010 WBES. We then list-wise delete the variables with missing values, leaving 2047 observations. The basic model closely follows variables 26 included in recent work assessing the determinants of productivity in the LAC region using WBES data (Arias Ortiz et al. 2014 ). We test the model using least squares (LS) and quantile regression techniques.
The basic model is: where the dependent variable is the log of labor productivity as measured by the fi rm's total annual sales at the end of the previous fi scal year (in 2009) 27 divided by the number of permanent and temporary 28 full-time employees at the end of the same previous fi scal year. The fi rm size is the log of the fi rm's response to the number of full-time permanent employees three fi scal years previously (in 2007). Age is a dummy variable that takes the value of 1 if the fi rm is less than ten years of age. Foreign ownership is a dummy variable that takes the value of 1 if the fi rm reports that more than 10 % is owned 29 by foreign individuals, companies, or organizations. Export is a dummy variable that takes the value of 1 if the fi rm reports more than 10 % of its sales are direct exports. Website is a dummy variable that takes the value of 1 if the fi rm reports having a website. Human capital is a continuous variable for the percentage of the fi rm's employees that are reported to have at least a university degree. Biggest obstacle fi nance is a dummy variable that takes the value of 1 if the fi rm reports access to fi nance as its biggest obstacle. Biggest obstacle edu WF is a dummy variable that takes a value of 1 if the fi rm reports an inadequately educated workforce as its biggest obstacle. We include these two obstacles because they were the most frequently cited in the sample, at 342 for fi nance and 316 30 for workforce education. We include country dummies to account for country-specifi c effects, such as the number of ports indi-  Ortiz et al. ( 2014 ). In 2007, fi rm size, whether or not the fi rm had a website, and human capital were statistically signifi cant. On the other hand, the age of the fi rm, foreign ownership, and exporter were not statistically signifi cant. This could be due to the fact that fi rms in the Caribbean tend to be older, on average, than in the rest of LAC. Also, as shown in Chap. 3 , there are relatively small proportions of fi rms in each country that are foreign-owned and similarly relatively small proportions of fi rms that export. The access to fi nance obstacle is highly signifi cant and negatively correlated with productivity. The inadequately educated workforce obstacle is not statistically signifi cant, but this changes if the human capital variable is dropped from the equation, at which point it becomes statistically signifi cant. In Table 7.7 (in the Appendix) we present results for just the manufacturing sector, where we include capital per worker. The access to fi nance obstacle remains signifi cant in the LS regression, but loses signifi cance in the quantile regression analysis. By restricting the sample to manufacturing fi rms with available data on capital, the number of observations drops to 600 fi rms, which limits the statistical power. We therefore present it more as a robustness check. We note that some of the other variables respond as expected. Exporter becomes statistically signifi cant and having a website loses signifi cance, which would be consistent with the different nature of business in manufacturing versus services fi rms.
The estimates for the different quantiles above and in the Appendix are the result of a simultaneous quantile regression that was bootstrapped at the standard 100 repetitions. This means that, while the coeffi cients and the pseudo R-squared do not change when the regression is run again, the standard errors can change slightly and some of the variables that are on the cusp of signifi cance can change. This can also affect whether the differences between the quantiles are statistically signifi cant. 31 The results from the analysis should be taken as preliminary evidence that not only do the fi rm characteristics vary as you move from lower to higher productivity levels, but also that these characteristics may affect performance to varying degrees, depending on where the fi rms lie in the distribution of labor productivity. One interpretation could be that there are fi rms in the lowest productivity category in which variables such as lack of access to fi nance are truly prohibitive. However, there are also slightly more productive fi rms that remain in the lower half of the productivity distribution. Despite reporting access to fi nance as an obstacle, these fi rms do not actually perform differently from their counterparts in the same part of the distribution who do not report this as their biggest obstacle. As fi rms move into higher productivity quantiles, those fi rms reporting access to fi nance as their biggest obstacle are indeed under-performing relative to the other fi rms in their performance quantile who do not report access to fi nance as their biggest obstacle. The preliminary results corroborate the notion that the characteristics of a fi rm and the obstacles it faces can indeed vary by and relate differently to productivity.

C ONCLUSION
The goal of this chapter was to better understand the fi rms in the Caribbean using micro-data from the WBES and PROTEQin. Since the data defi cit has been acknowledged as a challenge for evidenced-based policymaking in the Caribbean, our descriptive data presents fi rm characteristics from several perspectives. We sought to distinguish whether, after calibrating by main product (or sector) and country, the relatively higher or lower productivity fi rms show different patterns with respect to the key characteristics linked to productivity. They do. The story of productivity in the Caribbean appears to be consistent with fi ndings outside the region. Therefore, this chapter should serve as a point of departure for further research to gain a deeper understanding of how the characteristics of the private sector in the Caribbean countries exacerbate (or perhaps do not exacerbate) stagnated growth. Preliminary evidence suggests that there is variation in the obstacles identifi ed by relatively higher or lower productivity fi rms and, perhaps more importantly, individual obstacles such as access to fi nance associate differently with productivity performance. This is an original contribution that has rich policy implications for those in the region who wish to tailor or nuance policies to different types of fi rms in their economies. If policymakers are interested in moving relatively lower productivity fi rms into the higher productivity realms, they should zero in on the particular obstacles that the relatively lower productivity fi rms face. If, on the other hand, policymakers are concerned about how to support their relatively higher productivity fi rms, they should focus on the subset of obstacles reported by those fi rms. Source : Authorsʼ elaboration using PROTEQin data Table 7.6 (continued)  the Barro andLee dataset (1950-2010), this variable is only available for fi ve of the surveyed countries, supporting what is mentioned at the outset of this chapter, namely that when it comes to commonly used indicators, many of the Caribbean countries suffer from a defi cit of data. 15. The questions are based on a fi ve-point scale from "not diffi cult" to "almost impossible." 16. Biene et al. ( 2008 ) offered empirical evidence that brain drain is detrimental. 17. For example, Gibson and McKenzie ( 2011 ) raised questions about the existence of brain gain and proposed ideas to frame the empirical analysis of a series of understudied aspects of the impact of highly skilled migration. 18. Diasporic entrepreneurs are defi ned as entrepreneurs who are tapping into the Caribbean or diaspora markets. 19. A total of 67 diasporic fi rms responded to an online survey. The sample was mainly gathered from Compete Caribbean's registered database of entrepreneurs that responded to the open call for the Caribbean Idea Marketplace (CIM). 20. Of the fi rms surveyed by PROTEQin 79 % cited the quality of education as a factor ranging from important to critical in causing skill shortages and almost 77 % cited a shortage in the number of local professionals trained by local institutions. 21. A shortage in local professionals trained by local institutions could also be affected by people who migrate away from the Caribbean to pursue educational opportunities elsewhere (Thomas-Hope 2002 ). 22. See Table 7.6 in the Appendix for complete results. 23. Hallward-Driemeier and Aterido ( 2009 ) pointed out that endogeneity remains a concern with the obstacle of access to fi nance. In other words, it may be precisely because these fi rms have low productivity that they experience access to fi nance as their biggest obstacle. That does not mean that they are not objectively experiencing this obstacle. 24. Goedhuys and Sleuwaegen ( 2009 ) confronted a skewed distribution of their dependent variable and, as they describe, classical regression approaches are a location shift where the covariates are conditioned to the mean and are interpreted as being associated with a shift in the mean, but not in the shape or distribution of the dependent variable. They used quantile regression because they were interested in the factors that stretched the tail of distribution and had a strong effect where the high-growth fi rms were located. 25. Since the mean could be distorted by outliers in the tail of the distribution. 26. This model differs from some of the other approaches used in other chapters of this book because of the necessity to focus on the services sector, which is extremely relevant in the Caribbean. Therefore, we choose to closely follow Arias Ortiz et al. ( 2014 )), who used an approach readily applicable to our analysis. For example, we initially do not include capital per worker in our model because capital is not available in the survey for the services sector. In order to check how the results would differ, we perform the same analysis for the manufacturing sector only. The results are presented in Table 7.7 in the Appendix. 27. Standardized in the dataset by defl ating all responses to 2009 US dollars. 28. Correcting for the number of months of the year during which the temporary employees were working. 29. In the sample there were 12 observations of the 2047 that were categorized in the dataset as state-owned enterprises; these 12 observations were included as domestically owned (and took a value of zero). 30. The next most frequently cited biggest obstacle (by 238 fi rms) was electricity. 31. In this case, the following variables are statistically different at the 95 % level among the quantiles: employment in 2007 (size of the fi rm), full-time employees with at least a bachelor's degree, and exports. The biggest obstacle being access to fi nance was signifi cantly different at the 90 % level.