The creative–destructive force of hurricanes: evidence from technological adoption in colonial Jamaican sugar estates

Abstract

While in the early part of the nineteenth-century Jamaica was one of the world’s leading sugar producers, the abolition of slavery, the flooding of sugar markets with cheap European beet sugar, and the equalization and finally elimination of sugar import duties across the British empire led to a need for more efficient ways to produce sugar. However, it has been widely noted that Jamaica sugar estates were late in adopting more efficient production techniques, arguably due to inadequate financing. This paper investigates what role the destructive forces of hurricanes may have played in inducing Jamaica to finally modernize its sugar production. To this end, we combine a geo-referenced exhaustive data set of Jamaican sugar estates with a measure of localized hurricane damage constructed from historical hurricane tracks over the period 1882 to 1930. Our econometric analysis shows that hurricane strikes increased the probability that a surviving estate upgraded its sugar processing technology, particularly when the price of sugar was high and the price of the other main exporting crop (bananas) was low. Additionally, while a government hurricane loan programme working through local loan banks did help plantations to adopt new machinery, this depended on the damage not being too large.

1 Introduction

Although Great Britain only colonized Jamaica in 1655, by the eighteenth century sugar had become the most profitable crop in the British Empire and Jamaica its largest producer in the (British) West Indies (Higman 1987), where the mode of apparatus was a heavy reliance on slave labour and the use of a fairly primitive technology for production, while also enjoying essentially a monopoly of the British market through protective measures (Schuyler 1918). However, while profits in the industry had always been extremely sensitive to the inherent volatility of the price of sugar, over the following century slave emancipation in 1834, the gradual reduction of preferential duties with Britain starting in 1846 and final elimination in 1874 (Lobdell 1972), and the domination of the world market by European beet sugar producers by the 1880s through rapid technological developments and subsidies caused a large rise in production costs and a sharp drop in sugar prices (Lobdell 1972; Mintz 1959). These fundamental changes in the nature of production, trade protection, and competition meant that sugar estates in the British West Indies for large stretches of the nineteenth century were repeatedly faced with two main choices: either exit the market or find ways to substantially lower their costs of production (Cumper 1954; Beachey 1957). In this regard, Beachey (1957) estimates that of the approximately 2200 estates operating in the British West Indies at the time of emancipation only between 750 and 800 existed by the end of the century. For those that remained in production, in many of the colonies greater cost efficiency was achieved by the amalgamation of smaller estates into larger ones to benefit from economies of scale and/or substantial capital investment in newer, more cost-efficient technologies (Lobdell 1972).¹

Once the jewel in the British crown of sugar production, Jamaica was particularly hard hit by the drastic changes in the market in the nineteenth century. Of the 600 estates in cultivation in 1832, by 1896 only 140 had survived (Beachey 1957).² Moreover, although some amalgamation did take place, the large distances between estates generally made this an unfeasible option in many areas of the island (Britain et al. 1897; Lobdell 1972). However, curiously many of those estates that were still in production by the end of the nineteenth century had still not undertaken much by way of technological upgrading to become more cost-efficient, as compared, for instance, to Guyana or Trinidad (Norman 1897a). In this respect, Sheridan (1989) has argued that the investments necessary to replace existing with more modern sugar production capital would have been substantial, and thus for the average estate in Jamaica, small and heavily indebted, not feasible through self-financing. Additionally, Jamaica at the time was characterized by a lack of access to foreign and public capital (Fe 1984) and an underdeveloped local banking sector that generally only provided small, short-term loans (Quigley 1989; Callender 1965; Huesler 2024).

It was not until well into the first half of the twentieth century that Jamaican sugar estates slowly became more cost-efficient in line with international standards (Fe 1984), so that there was a long period of coexistence of plantations with both up-to-date and outdated technology. While it has been suggested that the high sugar prices caused by World War I were the primary cause (Callender 1965), in reality adaptation was much more gradual, essentially implying a prolonged coexistence of technologically advanced and laggard sugar producers on the island. There may, however, have been another type of a large shock that played a role in the technologically inferior Jamaican sugar producers finally upgrading. More specifically, lying in the North Atlantic Basin, Jamaica is periodically subject to potentially very damaging hurricanes and these storms have been shown to historically have had large effects on sugar estates across the West Indies, resulting in significant losses by destroying sugar cane harvests and, more importantly, by damaging sugar production infrastructure (Mohan and Strobl 2013; Smith 2012; Mulcahy 2004; Schwartz 2015). By the early twentieth century, such destruction of vital capital stock arguably would have made the repair of outdated equipment no longer a feasible option for those Jamaican sugar estates that wanted to remain in production. This paper thus investigates whether hurricanes may have indeed acted as a creative–destructive force inducing technologically laggard Jamaican sugar producers to upgrade if they could afford to do so.

To motivate our empirical analysis, we, as Dye (2011) did for technological adoption in the Cuban sugar industry, refer to Salter et al. (1969) vintage capital model, which allows for sugar producers with heterogeneous technologies to coexist in equilibrium. More specifically, because technology is arguably capital embodied in sugar processing, the cost of purchasing new equipment will be subtracted from net per unit of production profits, unlike for the sunk cost of existing equipment. Thus, there can be a range of different vintages of technology in use, depending on the costs of financing, heterogeneities in variable costs, or different forecasts of future prices (Dye 2011). Related to this, Atack et al. (1980) show that more advanced technologies, such as steam engines, produced both more and cheaper power. In our context, since plantation owners generally were able to avoid damage to crops by timing harvesting to take place between January and May while hurricane season generally runs from June to November (Satchell 1990), we consider hurricanes as mainly inducing negative shocks to the processing equipment, resulting in its necessary repair if production is to continue. However, if the repair costs render per unit production unprofitable, then plantation owners may exit the market or instead choose to upgrade their technology, assuming it is not already at the frontier. Importantly, Bloch et al. (2011) shows, by combining Salter et al. (1969) analysis of capital-embodied technical change with Kalecki (1968) analysis of financing investment, that retained profits can induce such technological investment. In the context of Jamaica, Sheridan (1989) has also argued that higher sugar prices provided the means to finance any technological upgrading. Feasibly then, once there was damage to equipment from a hurricane it may only be when sugar prices had previously been high that self-financing of such an investment was possible for many estates.

The empirical analysis of our study rests on combining an exhaustive geo-referenced database of Jamaican sugar estates over the period 1882 to 1930 with a local measure of destruction due to hurricanes. Importantly, for each sugar estate we have detailed information on the technology used for sugar production, allowing us to follow the estate level of sugar-producing technology over time. To proxy hurricane damages, we combine historical tropical storm tracks with a wind field model in order to differentiate likely estate specific damages across space and time. With these data at hand, we first establish that hurricanes resulted in reduced sugar production and some estates to exit from the market. We then show that destruction increased the probability of surviving estates updating their capital stock, but that this crucially depended on having the finances to do so.

There already exists some quantitative evidence on the possibly beneficial effects of large environmental disasters in the past, such as for urban growth (Hornbeck and Keniston 2017) or innovation (Noy and Strobl 2022, 2023). More importantly, there are also a handful of studies that show that such events can induce surviving enterprises to upgrade to newer technologies after capital stock destruction. For example, for the case of the 1923 Kanto earthquake in Japan Okazaki et al. (2019) provide evidence that local damages induced some firms to shut down 2 years later, while others increased the level of technology, as proxied by horsepower of their machinery. These effects crucially depended on firm size, which the authors interpret as a proxy of access to finance. Looking at firm survival and post-performance after the 1959 Isle Bay Typhoon, Okubo and Strobl (2021) found rather heterogeneous experiences across sectors in Japan, in that in some the flooding caused by the storm decreased survival but also increased the value of the capital stock in survivors, while the opposite was true in others.

Our paper contributes to the existing literature on a number of fronts. Firstly, from a methodological point of view our data set spans enterprises producing essentially a fairly homogeneous product (sugar) over nearly 50 years. This is also a period in which 7 damaging hurricanes struck Jamaica, providing us with a setting of a series of quasi-experiments, rather than one single event, with which to explore the question at hand. We additionally have precise measures of sugar production technology in that we can identify the type of machinery employed and hence can clearly identify what constitutes upgrades. Secondly, while it has previously been shown that the development of financial markets have historically been an important factor for innovation and adoption,³ we here instead explore the role of self-financing in the absence of financial markets. In particular, we investigate how the price of the product produced (sugar) as well as the price of the main other, and relatively capital non-intensive, agricultural product at the time (bananas) (Callender 1965), determined estates’ choices and timing of whether to upgrade. Finally, we explore how in such a limited capital market setting government intervention through providing loans may have been important for technological adoption.⁴

The remainder of the paper is organized as follows. The next section provides an overview of the historic background and general setting of Jamaica sugar production during our sample period, followed by an outline of the data in Sect. 3, methodology and the econometric analysis in Sect. 4. The last section contains a brief conclusion.

2 Historical background

2.1 Jamaica’s sugar history

Sugar production in Jamaica is estimated to have started around the beginning of the sixteenth century, when the island was still under Spanish rule (Woodward 2008). However, it only became an important economic actor after the British took control of the island from the Spanish in 1655, and by the seventeenth and eighteenth centuries, under the British plantation system, Jamaica became the major producer and leading exporter of sugar in the British Caribbean (Deerr 1950). An important feature of the sugar plantation system was its heavy reliance on slave labour. Thus, slave emancipation in 1838 constituted a particularly large cost shock. Given Jamaica’s relatively low population density, many former slaves moved away from the estates and were able to acquire their own land for agricultural production, and thus, labour supply was limited or unreliable and wages consequently higher (Cumper 1954). As a matter of fact, it has been estimated that wages of free labour accounted from anywhere between $\frac{1}{2}$ and $\frac{2}{3}$ of the total costs of producing sugar (Cumper 1954; Beachey 1957). One possible remedy would have been to become more cost-efficient by investing in newer production technology. However, absentee ownership was a common feature of Jamaican sugar plantations, where many absentee proprietors enjoyed extravagant lifestyles, financed by additional mortgages on their often already heavily indebted plantations, hence resulting in limited access to finance for most plantations (Reid 2016; Mintz 1959; Mulcahy 2004; Norman 1898b). Moreover, they were generally not able to obtain loans from locally based banks,⁵ making it thus difficult for many owners to invest in modern machinery to increase production after emancipation (France 1984; Britain et al. 1897; Lobdell 1972).

Beginning in 1825, with the entrance of Mauritian muscovado sugar into the global market, Jamaican sugar planters also saw the near monopoly of British West Indian sugar colonies of the British sugar market slowly eroding. This culminated with the Sugar Duty Act of 1846, which introduced the gradual elimination of preferential duties by 1854 within the British empire. In 1884, the price of sugar further declined due to the competition from cheap German and Austrian-Hungarian bountied beetroot sugar, which flooded the European market (Lobdell 1972; Mintz 1959). Consequently sugar exports to Britain fell substantially, although this was partly dampened by increased consumer demand from the US market (Lobdell 1972). This in turn led to the abandonment of unproductive lands and the restriction of cultivation on some plots (Handbook of Jamaica 1880). Falling, and uncertain, profits subsequently discouraged many planters from investing in modern machinery (Civil & Military Gazette 1882). As cane cultivation stagnated and global demand for bananas rose, there was a significant shift towards banana cultivation. The export of bananas more than quadrupled between 1880 and 1884 in Jamaica, where a number of sugar estates switched to banana production. As noted by Lobdell (1972), most of the shifting estates were also those unable to get the financial means to update their machinery for sugar production. In 1902, Jamaica’s sugar industry was further severely affected by the Brussels Sugar Convention, which abolished both direct and indirect bounties benefiting the export of sugar (Taylor 1909). This induced the passing of Law 43 of 1903, removing all sugar duties, so that sugar, rum, and molasses from other countries could be freely imported into Jamaica. It is nevertheless noteworthy that not all sugar plantations had decreased their production at this stage, as a few estates, like for example Bushy Park in the parish of St. Catherine, managed to increase their output by introducing state of the art technology (The Jamaica Gleaner 1896).

The end of the 1910s was marked by a significant increase in sugarcane cultivation and high prices for all crops, but especially for sugar, most likely due to the collapse of European beet sugar production and the sudden stark increase in British sugar tariffs, which had been gradually reintroduced in 1901, during WWI. However, the subsequent fall of the price of sugar from £75 to £12 per ton in 1921 after dismantling of the tariffs had a devastating effect on sugar estate owners, particularly those that had used the profits due to the prior high prices to upgrade machinery. More specifically, there had been a “... general conviction that high prices would last another season, and this included many planters to improve facilities and their factories. This meant the buying of machinery at abnormally high prices. Then, too, the planters did not sell the whole of their sugar on the top of the boom. This meant that planters instead of getting as they hoped £100 for a ton of sugar, they received anything from £40 to as low as £15 a ton. On the top of this came the bills for machinery to be paid when the planters’ means were exhausted” ( The Jamaica Gleaner 1921, p. 3).

The trends in Fig. 1 demonstrate the changes just described in terms of Jamaican sugar exports, share of sugar in total exports, and the price of sugar. As can be seen, the total quantity of sugar exported rose steadily from 1700 to the early part of the nineteenth century, and then experienced a prolonged decline until the early part of the twentieth century when there was a stark rise. In terms of the share of the value of sugar in total exports, one notices also a prolonged decline, and, although the share increases substantially as the per unit market value rose in the late second decade of the 20th, never quite reached the heights of constituting over 50 per cent of total exports witnessed prior to second half of the nineteenth century. One can also see that this decline in relative exports is also reflected in Jamaica’s share of the world market, where this fell steadily from the mid-19th until it plateaued in the beginning of the twentieth century.

Fig. 1

Evolution of the Jamaican Sugar Market. Notes: This graph illustrates how the sugar exports (1000 tons) evolved over time (lower). The two vertical lines show the start and end year of the observed period. The upper panels show the share of sugar from Jamaica’s total exports (left) (The underlined data comes from Board of Trade (1883), The Governor’s Report (1889), Blue book of Jamaica (1838), Blue book of Jamaica (1842), Blue book of Jamaica (1846), The Governor’s Report (1892), The Governor’s Report (1892), The Governor’s Report (1896), The Governor’s Report (1925), The Governor’s Report (1927), The Governor’s Report (1930), Handbook of Jamaica (1886), Handbook of Jamaica (1908), Handbook of Jamaica (1916), Handbook of Jamaica (1921) and from Handbook of Jamaica (1923).), the price per ton of sugar in £ (middle), which has been deflated with a composite price index from O’Donoghue et al. (2004) and the share of sugar produced in Jamaica relative to the total global sugar production (Data from (Deerr 1950, ][pp. 490–491)). The grey shaded area is the study period in each panel

2.2 Sugar production technology in the late nineteenth and early twentieth centuries

The first step in sugar processing was the extraction of cane juice, which was performed by a mill. In the earlier years, mostly single roller mills were used to crush the cane and extract the juice (Beachey 1957). However, over time improvements like applying multiple, heavier, or hydraulic rollers became crucial in maximizing the amount of extracted cane juice (Beachey 1957). The motive power of the mill varied as well, where the simplest were the animal (mostly cattle) and water or wind powered mills (when these power sources were available), and finally steam mills (France 1984). The first attempt to use steam power was in late nineteenth-century Jamaica, roughly a decade after the first successful application in Cuba in 1797 (Deerr 1950). Even after the successful application in Cuba, the spread of steam engines in Jamaica was still very slow despite first being used in 1808 (France 1984). For instance, in 1871 almost all Jamaican Sugar Mills were moved by animal mills and only one out of over 5600 Sugar Mill was moved by steam power (Minutes of the Legislative Council 1871). In the late nineteenth century, upgrading mills, however, became much less costly as the production of steam engines was standardized, which also led to a higher adoption rate of steam mills on sugar estates (Tann 2016). As a matter of fact, by 1900 almost all estates were using steam mills (Cumper 1954) (Fig. 2).

Fig. 2

Jamaica’s Share of English Caribbean Sugar Exports. Notes: This graph illustrates Jamaica’s share of sugar exports compared to the other sugar exporting Caribbean colonies in the British Empire, such as Barbados, Antigua, Montserrat, the Virgins, Grenada and the Grenadines, St. Vincent, St. Lucia, Dominica, Trinidad, Tobago, Nevis, and St. Kitts. The underlying data comes from Deerr (1950). The grey shaded area is the study period

After extraction, the cane juice was heated to 140 degrees Celsius, and a clarifying agent was added. This was followed by evaporation and concentration, during which the sugar was purified and glucose was obtained (Beachey 1957). Evaporation of the syrup traditionally took place in a range of copper boilers over an open fire (Beachey 1957). The major problem with this earlier technology was that evaporation required very high temperatures. Consequently the invention of the vacuum pan became one of the most crucial advances in the sugar industry, as it was able to boil syrup, in vacuo, at much lower temperatures than before (Beachey 1957). Further improvements in the vacuum pan included triple, quadruple and multiple effects, which made the entire process much more efficient, as less energy was needed to evaporate sugar (Deerr 1950; Minutes of the Legislative Council 1887; Fuga 1961). For example, it has been noted by Beachey (1957) that the vacuum pan could have reduced the costs of sugar production by as much as 50 per cent. Nevertheless, the adoption of the vacuum pan was generally slow in that it was first introduced in 1818 in Europe and only made its first appearance in Jamaica some 25 years later and remained uncommon for much of the nineteenth century (Deerr 1950). The fact that, in contrast, the adoption of the vacuum pan took place relatively earlier in some other colonies, such as Guyana or Trinidad, was likely due the fact that sugar estates in these tended to be larger and less indebted, and thus could afford the considerable financial outlay required (Beachey 1957). Additionally, the relative labour shortage in Jamaica, as many former slaves moved away from the plantations, may have played a role (Beachey 1957; Engerman 1983).

After evaporation, the molasses was separated from the sugar crystals. Originally this was done by natural drainage, which required two weeks or more (see Beachey 1957). The introduction of the centrifuge greatly accelerated this process and only took a few hours (see Beachey 1957; Fuga 1961). Additionally, they were easy to install in existing machinery, and second did not require any new skills (Cumper 1954; Fe 1984). Moreover, with a centrifugal present the most could be gotten out of a vacuum pan (Beachey 1957).

Importantly, innovations in the processing of sugar, like the introduction of vacuum pans and centrifuges, also made sugar production less labour intense (Satchell 1989). There was a clear awareness of this in the region at the time as “[t]he general consensus of opinion seems to show that sugar planting in the West Indies may again become a profitable industry. At present, however, things are in a very slovenly state. Methods of cultivation have been improved, but the old-fashioned machinery is inadequate to cope with the advance made in the sugar trade. Many of the old mills only got 60 per cent out of the cane, but with modern machinery, it would possible to get quite 80 per cent., if not more” (Bath Chronicle and Weekly Gazette 1901, p. 2). Nevertheless, the adoption of the new processing technologies was slow. For example, in the case on Worthy Park Estate in Jamaica acquiring a vacuum pan was recommended as early as 1875 (Craton and Walvin 1970), but was instead postponed until 30 years later in 1906 (Craton and Walvin 1970). Improvements of the underlying process were much more common after 1900, where a number of estates are known to have undertaken these improvements when sugar prices were high and some of the revenue generated could be added to cash reserves (Callender 1965). For example, during the period 1919 to 1928, sugar estates used their savings to increase production and invested over £1,300,000 in new equipment (Callender 1965).

Investment in sugar production technology by estate owners was not the only strategy for combating falling profits in the industry of the late nineteenth century. Another solution proposed was to centralize sugar milling and processing at a central factory, as had been seen to be successful in, for instance, the French West Indies. (Beachey 1978). This was an option strongly supported by the 1897 West Indian Royal Commission, specifically set up to examine the condition and prospects of the British West Indian sugar colonies. For example, in the parish of St. Catherine in Jamaica in 1882 it was argued that a central factory could purchase “[...] sugar canes from neighbouring proprietors and lessees and manufacturing the same into Sugar and Rum [...]” (The Budget 1882, p. 3). That is, as smaller estates were not able to process sugar inexpensively on their own, they could benefit from a central factory that purchased their cane at moderate rates (Beachey 1978). In order to encourage the construction of central factories, the colonial government in Jamaica in response passed a law in 1902 that guaranteed the interest on any capital manufacturing or preparing any products of the island (Callender 1965). Nevertheless, at the beginning of the twentieth century there were still only a few central factories in St. Thomas and St. Catherine (Handbook of Jamaica 1880), and it is only in starting in the 1930s that sugar production was dominated by central factories (Fe 1984).⁶ There were arguably a number of reasons for this. For example, as noted by Beachey (1957), only few sugar areas in Jamaica were believed to have enough cane grounds to make it profitable, where a minimum of 38 square miles was believed to be necessary.⁷ Additionally, there was as a lack of suitable transport infrastructure in terms of both roads and railways. More specifically, central factories would have required a sufficient railroad network to transport the sugarcane quickly from the plantations to the factories. However, while there were some railroads in Jamaica this was not sufficient to support central factory production and metropolitan investors were unwilling to invest in any necessary expansion (Green 1973). Moreover, even by 1920 zero per cent of roads in Jamaica had been paved (Maunder 1954). Other reasons for the lack of investment in central factories include a lack of financing available (Britain et al. 1897), worries by the merchants who held mortgages on the plantations that central factories would have prior claim on the sugar produced, as well as concerns by sugar plantation owners that relying on central factories would affect their image of independence (Beachey 1957; Green 1973).

2.3 Hurricanes and sugar production

The sugar industry in the Caribbean has a long history of being particularly vulnerable to damages induced by hurricanes (Mulcahy 2004; Schwartz 2015). In this regard, using colony sugar export data Mohan and Strobl (2013) estimate that a damaging hurricane could completely wipe out sugar production in the smaller islands, and substantially reduce that of the larger islands. In Jamaica, for example, the average reduction ranged between 20 and 25 per cent, depending on the proxy of damages used.

With regard to the period examined here, there were several damaging hurricanes in Jamaica that are well documented to have affected sugar production. For example, in 1880 Jamaica was hit by a hurricane that caused significant damage to some sugar estates and their machinery, as described, for instance, by The Saint Christopher Gazette and Caribbean Courier (1880): “Along the shore, the Estates of Lyssons, Retreat, and Leith Hall have had some of their buildings unroofed. At [R]etreat, a large chimney fell, carrying the roof of Engine and Mill House with it and damaging Machinery” (The Saint Christopher Gazette and Caribbean Courier 1880, p. 2). The agent from the estate of owner Alexander Crum Ewing described the destruction to the estates as follows: “The damage done is very great. [...] At Dawkins Caymanas the boiling-house roof is quite destroyed, half of the trash-house roof, the still-house chimney was blown from its foundation right across the tank and deposited on the roof the distillery, 25 feet off, showing the force of the wind. [...] Mr Verlay is one of the worst sufferers, the damage done to ‘Mona’ he estimates at £3000. Then the mills are much injured, 14 out of 18 oxen destroyed-the roof of the mill fell down upon them-and his wharf is washed away” (Glasgow Weekly Herald 1880, p. 8).⁸

A few years later in 1886, Jamaica witnessed a tremendous storm that left the sugar industry in the parish of St. Thomas in dire straits (Minutes of the Legislative Council 1886). Some estates, such as Ewings Caymanas, used the opportunity to upgrade their machinery (The Governor’s Report 1885).⁹ Almost a decade later, in November 1912, there was a heavy storm which caused damage to the parishes of Portland’s and St. Mary’s cultivation. Some days later, a different hurricane ravaged the western parishes St. James, Hanover, and parts of Westmoreland, resulting in a 30% loss of sugar production (The Jamaica Gleaner 1913). These hurricanes caused extensive damage, as in the case of Fred L. Clarke’s estate, where the storms destroyed large portions of crops as well as the estate’s buildings and machinery. Importantly, hurricanes such as these often forced planters to turn to creditors as they set out to rebuild their capital infrastructure (Mulcahy 2004).

2.4 Financing

2.4.1 Cost of advanced sugar technology

The cost of upgrading sugar processing technology was considered to be generally prohibitive, particularly in view of the volatile and uncertain market price of sugar (Britain et al. 1897). In terms of the cost, data from Norman (1897b), for example, show that in British Guiana the average sugar estate invested over £14,000 in new machinery between 1882 and 1895. Roughly 15 years later, George Carrington, an estate owner from Barbados, mentioned that updating machinery of an estate costs about £10,000 (Norman 1897a).¹⁰ Given that data on production and production costs on a select number of sugar estates in 1897 collected by the Sugar Planters Association (Britain et al. 1897) and sugar and rum prices in that year from Blue book of Jamaica (1897) suggests that the average annual profit for estates without a vacuum pan and centrifugal was £542, which implies that if such an investment were to be solely financed by savings at 1897 prices then this would take at least 20 years to repay. One should note that the potential increase in profits in upgrading suggested by the same data would be sizeable, as the average profit in estates with updated processing technology in the sample was £3,504, while the average per acre profit was more than double (£6.44 versus £3.07) and the average profit per ton of sugar was nearly double (£6.02 versus £3.34).

2.4.2 Access to finance

As a large proportion of the sugar estate owners in Jamaica were absentee and heavily indebted, it was difficult for them to invest in new machinery or to expand cultivation (Sheridan 1989).¹¹ But even for landowners without large debts, who often put all their assets into their estates, raising additional money for new technologies proved difficult (Minutes of the Legislative Council 1873). In terms of private banking in Jamaica, the British Colonial Bank held a monopoly until 1890, when a few Canadian banks entered the market (Quigley 1989). However, the Colonial bank was not able to provide long-term investment capital, as it was prohibited from lending against property (Monteith 2003; Lobdell 1972), while the private banks were generally risk averse, only granting small, if any, loans. For example, in 1906 the Bank of Nova Scotia loaned only 15% of its deposits (Quigley 1989). Hence, arguably the banking system was not able to provide estate owners with the loans financing for large investments.¹² As a matter of fact, frustration in this regard was often voiced by the plantation owners. For example, in reporting to the West India Royal Commission one estate owner stated that “There is practically no credit [...] and I cannot borrow any money” ( Norman 1897a, p. 256).

2.4.3 Government financing

The Jamaican government generally did not provide direct financing to sugar estates except on three notable occasions over the period under consideration.¹³ First, in response to the devastating hurricane of 1903, the Hurricane Loan Law of 1903 was passed, providing loans in aid of restoration of cultivation due to the hurricane (Hoyte 1969). However, these loans were limited to no more than £3 per acre (Handbook of Jamaica 1904) and thus even for the largest sugar producer at the time ($\approx$ 1000 acres) would have hardly provided sufficient funds to fully finance full investment in new processing technology, let alone for the average sized estate ($\approx$ 200 acres). In addition, the loans had to be repaid by 1905, i.e. relatively shortly after being received.

In response to the disastrous hurricane in 1912, the government established the Agricultural Loan Society Board with the Agricultural Loans Societies Law in 1912 (Laws of Jamaica 1912). The Agricultural Loan Society Board was entitled to provide some money to the newly installed Agricultural Loan Societies (see Minutes of the Legislative Council 1912). The Agricultural Loan Societies’ shares were owned by its members, i.e. people and enterprises connected to agriculture or agricultural trade, which in turn entitled them to receive loans of up to £200, if either the 1912 hurricane or the drought during that period caused damage to their cultivated land (Laws of Jamaica 1912). Again, this amount was unlikely to be enough to finance technological upgrading in sugar processing. However, 1 year later, in June 1913, £30,000 was allocated by the agricultural loan societies board for the relief of the sugar industry in Westmoreland and Hanover, who suffered losses from the hurricane (see Minutes of the Legislative Council 1913, 1914). In addition, under Laws 36 and 37 of 1912, approximately £50,000 was appropriated to People’s Co-operative Loan banks. These local agricultural loan banks were originally established in 1905 and were intended to provide small farmers with longer-term loans, but had nearly all failed until the passing of the Agricultural Loan Societies Law of 1912, which established a board to support the Agricultural Loan Societies, through which public funds then could be made available to the local loan banks to lend to its members (Callender 1965). As noted by Callender (1965), the subsequent primary recipients of such loans were the owners and lessees of sugar crops damaged by the hurricane. For instance, of the £50,000, 40% went to five local loan banks in St. Mary and 25% to six banks in Clarendon, St. Catherine, and St. James, whereas only a very small proportion, i.e. £115, went directly to cane farmers located in Hanover (see Minutes of the Legislative Council 1914). Importantly, as these local loan banks were able to provide fast loans after natural disasters, like after the 1912 storms, they became valuable institutions (see Minutes of the Legislative Council 1914). Moreover, the government decided to provide direct loans of £43,646, from which over half of it went to Vere in the south, which is now incorporated into Clarendon Parish (see Minutes of the Legislative Council 1914). 3 years after issuing the loans, almost half (The Governor’s Report 1916) and 10 years later most of the loans were repaid (The Governor’s Report 1922).

A few years later, when sugar prices fell rapidly in 1920, many estates again found themselves in severe financial difficulties, the Jamaican government decided to help the sugar industry with the Sugar Industry Aid Loan Law of 1921, which provided a total amount of loans of up to £400,000 to the planters (see The Jamaica Gleaner 1921; Minutes of the Legislative Council 1924). As prices continued to fall in the 1920s, the Legislature of Jamaica additionally introduced an export bounty of £2.00 per ton of sugar in 1929, which led to a massive increase in sugar production in 1930 (The Governor’s Report 1929, 1930). The introduction of the export bounty was crucial, as it prevented many sugar estates from bankruptcy which would have destroyed the millions invested into updating the machinery and extending cultivation (Minutes of the Legislative Council 1930).

Finally, one should note that there was generally little support from the British colonial government itself for the technological upgrading in the sugar industry in general (Lobdell 1972), or providing relief in response to damages induced by hurricanes. With regard to the latter, Webber (2018) argues that “[...] colonial responses were ad-hoc, and fraught with anxiety due to the need to respond to shortages of food and materials” (Webber 2018, p. 0).

3 Data

3.1 Estate data

Data on Jamaican sugar estates is taken from the annual Sugar Estates in Cultivation tables collated by the Institute of Jamaica in the Handbooks of Jamaica, which were first published in 1881 and provide information on the production, as well as various aspects of ownership, of all sugar estates in cultivation in Jamaica¹⁴, as well information on their ownership and management, acreage, manufacturing technology, and output. We digitalized the data up to 1930.¹⁵

The Sugar Estates in Cultivation tables explicitly state the estate’s name, owner’s name, and, if applicable, the managing attorney, as well as its parish of location. This allowed us to match estates’ observations over time. Linking the estate name with information collated by the University College London Department of History’s Centre for the Study of the Legacies of British Slavery database¹⁶ also enabled the identification of the exact geographic location of each estate, as well as their age. From these data, we created a dummy for whether the estate was managed by an attorney (ATTD), the age of the estate (AGE), and the number of estates owned by each owner (OWNERSUM). The inclusion of OWNERSUM and ATTD as a control variables stems from the fact that some estate owners owned multiple estates in order to benefit from economies of scale (Higman 2008). It is expected that estate owners with multiple estates would be more productive and have greater financial resources to upgrade machinery. However, owning multiple estates is associated with higher costs for estate management. As a result, owners often hire attorneys to manage their estates. A professional attorney who is able to introduce innovations can have a positive impact on the estate business (Sheridan 1957, 1960). Nevertheless, the presence of an attorney may also be an indicator that the owner was an absentee. AGE was introduced as a control variable due to intensified agriculture reducing soil quality (Cherubin et al. 2016; Cheesman 2004) and, consequently, sugar production on the estate.

In terms of sugar production, the tables provide the extent of area (in acres) under sugar cultivation and the amount of sugar (in hogshead¹⁷) produced. From this, we calculated the annual sugar production per acre (HPRS), expressed in terms of its inverse hyperbolic sine in order to take account of occasional zero production years.¹⁸ With regard to milling technology, there is information on the type of mill employed, where there is a distinction between wind, water, cattle, and steam driven mills. Since the steam mill is the most up-to-date technology of these, but at times is used in conjunction with other mills, we simply define a dummy (MILL) of when a steam mill is employed. With regard to evaporation technology, the data allows one to identify when aspinal, wetzel, or vacuum pans were used, and we thus create a dummy (TECHV) for the use of the latter, most advanced, technology.¹⁹ For the separation of the molasses from the sugar crystals in the processing of sugar by estates, one can also determine whether the more modern centrifugal apparatus was used, and we accordingly created an indicator variable (TECHC). Additionally, we generated a dummy for when both a vacuum pan and a centrifugal was in use (TECHCV) and, by examining the previous year’s data whether either of these technologies or both of them were introduced for the first time, generating corresponding dummies fTECHV, fTECHC, and fTECHCV.

A sugar plantation’s exit from the sugar-producing market was identified when it no longer appeared in the tables.²⁰ The information given also allowed us to identify when an estate no longer produced sugar itself, but instead sent the canes to a central factory. Unfortunately, we have no information as to the production of the central factories itself, but do have the acreage planted from the supplying estates. We created dummies for both market exit (EXIT) and central factory production (CF) for the year prior to these.

3.2 Hurricane 1912 loans

The Governor’s Report (1926) compiled a list of all the local agricultural loan banks that extended loans under the Hurricane Law of 1912 and the corresponding total amount given, where there were a total of 38 of such lending establishments. We proxied the most likely relevant loan bank for each estate by identifying the one closest to the estate, using the geo-coordinates of the estate and the latitude and longitude of the town of the bank’s location.²¹ Loans per estate were then simply calculated as the total amount of loans of the corresponding bank divided by the number of estates for which the bank was the nearest (LOAN). One should note that this variable thus is constructed to vary across estates even within parishes.

3.3 Hurricane damages

In order to construct a proxy of damages to sugar estates due to hurricanes we use the historical storm tracks from the National Hurricane Center’s North Atlantic Hurricane database (HURDAT) as inputs into a wind field model, as in Strobl (2012). HURDAT consists of the tracks of all known tropical storms in the North Atlantic since 1855, providing the six hourly position of the storm eye as well as the maximum wind speed at each point, as derived from ship sightings.²² We interpolate these to hourly data observations. Using these within a tropical storm wind field model allows us to predict the maximum wind speed experienced during a storm’s lifetime at any point relative to the eye of the storm, and we do so for the location of each Jamaican sugar estate for each storm. Following Strobl (2012), damages are assumed to have a cubic relationship to wind speed due to energy dissipation reasons (Emanuel 2005), and that damages occur once wind speeds reach 119 km/h, i.e. Saffir–Simpson Scale Level 1. As the consequent value of this variable (H) was large, we normalized it by $10e^{-06}$. Additionally we created its mean ($\overline{H_{i}}$) and standard deviation ($H_{i}^{\sigma }$) for the period prior to 1880 to capture its estate specific pre-sample period distribution.

3.4 Price of sugar and bananas

In order to proxy the prices of sugar and bananas, denoted as PS and PB, respectively, we use the unit value as taken from the Jamaican Colonial Blue Books. This is constructed by dividing the value of exports of each good by its quantity.²³ As these are in different units (tons for sugar and bunches for bananas), we generated their anamolies, i.e. deviations from the mean normalized by the standard deviation.

3.5 Summary statistics

Figure 3 shows the spatial distribution of the 203 estates in our data, where the size of the dots indicates the relative number of years an estate was in the sample. Accordingly, most of the estates were concentrated in the western parishes of Westmoreland, Hanover, Saint James, and Trelawny. The majority of estates remained for at least 20 years in the sample, but with large variations.

Fig. 3

Distribution of Estates. Notes This graph illustrates the distribution of the sugar estates in Jamaica during the observed period. The colours indicate the number of years, a specific estate was part of the sample (colour figure online)

Summary statistics of all our variables for the estates are provided in Table 1. As can be seen, the average (inverse hyperbolic sine) sugar production per acre is 0.92, corresponding to 2.4 hogshead per acre in levels, but with a variation of 80 per cent of the mean. Four per cent of the observations are estates producing right before they exit market, i.e. 173 of plantations ceased production within our sample period. The average age of estates is 131 years, where 68 per cent of these are run by attorneys and most plantations are part of a single ownership structure. The average use of steam mills is 86 per cent. For processing technology, the use of centrifugals (0.72) over the sample period is larger than that of a vacuum pan (0.48), where only 38 per cent of the observations constitute estates using both. In terms of first time adaptation of the processing technologies,²⁴ their low value but high standard deviation indicates that much updating has taken place over the period in question. As a matter of fact, as shown in Fig. 4, the number of estates that already had the technology in place at the beginning of the period was only 27 and 45 per cent for vacuum pans and centrifugal, respectively, and 7 per cent for both. This rose to 100 per cent in terms of the centrifugal and 93 per cent for vacuum pan by 1930. One should note, however, that some of the increases in shares were due to exits of non-users and not just first time adoptions. The higher share of vacuum pan compared to centrifugal technology is likely due to the fact that the former required less skilled labour to operate (Cumper 1954).

Table 1 Summary statistics

The use of the historical storm tracks within our wind field model and damage function identified 7 damaging storms in Jamaica during our sample period, namely in 1886, 1895, 1903, 1912, 1915, 1917, and 1928.²⁵ The variation of H for these storms is more than double its mean, which is in large part because for most years for most estates the damage due to hurricanes is zero. The average annual value of H across estates are depicted in Fig. 4. Accordingly, the highest average destruction was in 1903, followed by the storm in 1912. Comparing the trends in technology adoption to the occurrences of hurricanes any evidence of an increase in the share of the technologies, singly or together, seems rather mixed. One of the possible reasons may be that upgrading after a hurricane was conditional on available financing. For example, after the 1915 hurricane, when prices would have been on a rise due to WWI, the share of estates employing vacuum pans increased.

Fig. 4

Share of Estates Using Modernized Sugar Processing Technology. Notes This graph illustrates the share of existing estates utilizing centrifugals, vacuum pans, or both. H is the hurricane damage index described in Sect. 3

The average loan amount from local banks in 1912 (LOAN) was £890, but with a standard deviation of 1850, where the largest implied amount was slightly over 10,000 and the smallest less than 10. Finally, the statistics for PS and PB show that the anamolies in banana prices were larger than those of sugar.

4 Econometric analysis

As a theoretical framework underlying our econometric analysis, we follow Dye (2011) approach to technological adoption in the Cuban sugar industry and refer to Salter et al. (1969) vintage capital, where plantation owners, once their estate’s capital stock is damaged by a hurricane, may continue as is, or, if the status quo is rendered unprofitable by the damage may either exit or instead choose to upgrade their technology, assuming it is not already at the frontier. To put this theoretical framework to the data we undertake a number of estimations. First, we seek to confirm that our hurricane proxy accurately captures damage done to sugar estates by examining its impact on sugar production. Secondly, we estimate the effect on the aforementioned choices for an estate once it is damaged, namely to either exit the market, remain as is, or, if this has not already been done, upgrade its technology.

4.1 Impact on sugar production

The impact of hurricanes on sugar production (per acre) is estimated with the following linear model:

$$\begin{aligned} PRS_{it} = \alpha + \beta _{H} H_{it-1} + \beta _{{{\textbf{X}}}} {{\textbf{X}}}_{it} + \lambda _t + \mu _i + \epsilon _{it} \end{aligned}$$

(1)

where PRS is the inverse hyperbolic sine value of sugar (hogsheads) per acre for estate i in year t. H is the hurricane damage index described in Sect. 3 measured at $t-1$ since sugar harvesting season is generally in January to May and hurricane season May to November. ${{\textbf{X}}}$ are a base set of control variables, including an estate’s age (AGE) and its squared value (AGE$^2$), a dummy for whether the estate is managed by an attorney (ATTD), the number of estates the owner possesses (OWNS), and whether the estate runs on as steam mill (MILL). $\lambda$ is set of year fixed effects, while $\mu$ are a set of estate fixed effects and $\epsilon$ is the usual error term. Standard errors are clustered at the estate level.

One should note that arguably the estimated coefficient on $H_{it-1}$, i.e. $\beta _{H}$, is exogenous and thus can be interpreted strictly causally. More precisely, while one may worry that estates make their location and other decisions with the knowledge of the local risk of hurricane wind exposure in mind, after controlling for the local distribution of such potential damages with estate fixed effects $\mu$, one is left with unpredictable, random realizations from this wind exposure distribution. We can thus confidently interpret any estimated effect of H in 1 as causal.²⁶

The results of estimating Eq. (1) including H at time $t-1$ are shown in the first column of Table 2.²⁷ As can be seen, hurricane wind exposure acts to significantly reduce per acre sugar production. If one takes the average nonzero value of H, i.e. 3.4, then this implies a change in the elasticity of PRS of $-$13.6 per cent. In the second column of Table 2, we experimented with using a higher damage minimum threshold value, i.e. 178 km/h for H. However, as can be seen the coefficient is substantially smaller and insignificant, indicating that such a functional form is not able to capture the damages incurred in sugar estates due to hurricane wind exposure. We also investigated in the third and fourth columns whether there might be lagged effects at $t-2$ and $t-3$, respectively, but their estimated coefficient are insignificant and this does not change the coefficient on H at t noticeably.

Table 2 Impact of hurricane damage on sugar production

In terms of the other control variables (${{\textbf{X}}}$), one finds that only AGE, in an inverted u-shaped manner, significantly predicts sugar production. More precisely, as an estate ages there is first a reduction in production, but at a decreasing rate. Although in principle this decreasing rate would ultimately result in an increase in production, the rate of decrease, i.e. the coefficient on ${\text{AGE}}^2$, is extremely small, suggesting that such a turning point (355 years) is well beyond any observed lifetime in our sample. The ultimately decreasing relationship may be due to soil exhaustion, where only older and established states have sufficient experience or financing enough to use techniques to counteract this erosion of quality of soil. However, one should note that the youngest estate in our sample is 64 years and thus that this relationship is estimated based on estates that have been in the market for many decades already. The fact that owning more than one estate does not translate into productivity gains is perhaps not surprising as it has already been noted earlier that one of the reasons why establishing central factories to merge the processing of cane across estates was not appropriate for most of Jamaica was because of the large distances between estates. A similar reasoning may be applied in terms of not finding any economies of scale for multiple estate ownership. The result that attorney management does not have a detrimental effect on sugar production has already been shown by Higman (2005) for Jamaica and for Saint Vincent by Smith and Forster (2018), although only for the first half of the nineteenth century.

4.2 Impact on market exit and switch to central factory production

Figure 5 shows the number of sugar-producing estates over our sample period. As can be seen these fell considerably from 176 to only 30 over our sample period. To explore whether hurricane damage induced estates to exit the processing market, broadly defined as complete abandonment or shifting to other types of production, such as bananas, we specify a simple probit model²⁸:

$$\begin{aligned} P\Bigl ( \hbox {EXIT}_{it} = 1 \Bigr ) = \Phi \biggl ( \alpha + \beta _{H} H_{it-1} + \beta _{{{\textbf{X}}}} {{\textbf{X}}}_{it-1} + \beta _{{{\textbf{Z}}}} {{\textbf{Z}}}_{i} + \gamma _K + \lambda _t + \epsilon _{it} \biggr ) \end{aligned}$$

(2)

where EXIT is the indicator variable of whether the estate is no longer in the market at $t+1$, P is the probability that EXIT takes on the a value of one, and $\Phi$ is the cumulative distribution function of the standard normal distribution. Thus, Eq. (2) is a nonlinear model so that one can no longer control for estate fixed effects that would capture all sugar estate time invariant unobservables. In particular, one might be worried about not taking account of the local hurricane damage distribution, thus rendering the causal identification on H questionable. Thus, we, in addition to the longitude (LONG) and (LAT) of the estate and a set of parish dummies ($\gamma$), also control for the mean ($\overline{H_{i}}$) and standard deviation ($H_{i}^{\sigma }$) of H as measured prior to our sample period, i.e. for 1855–1880, as denoted in ${{\textbf{Z}}}$. Standard errors are clustered at the estate level, and all coefficients are reported as marginal effects.

Fig. 5

Number of Sugar-Producing Estates. Notes This graph depicts the number of estates which are planting and processing sugar. H is the hurricane damage index described in Sect. 3

The results of estimating Fig. 4 are provided in the first column of Table 3. Accordingly, there is no apparent effect of hurricane damage on the probability that an estate exits the sugar production market. However, when we include H’s lagged value in the second column one finds a significant positive impact on EXIT, where this exit inducing effect does not last beyond $t-2$ (Column 3). The estimated coefficient on $H_{t-2}$ suggests that the average nonzero hurricane damage increases the probability of an estate ceasing production 2 years later by about 6.8 per cent, while the largest observed impact (7.4) over our sample period was 14.8 per cent.

Table 3 Impact of hurricane damage on estate exit and central factory production

Figure 6 depicts the share of sugar estates outsourcing processing to a central factory as well as their share in total sugar acreage. Accordingly, both the number as well as their acreage increased substantially in importance, where by 1930 these constituted nearly 20 per cent and 40 per cent, respectively. To see whether hurricane damage may have resulted in increased outsourcing, we re-estimated Eq. (2) but instead in terms of switching to central factory production. As can be seen, hurricane damage had no impact on switching to production to central factories, regardless of the time horizon considered. In terms of the other control variables, one finds that only the latitude and longitude are significantly significant. This may not be surprising since, as has been noted earlier, many plantations were geographically disadvantaged to produce for a central factory.

Fig. 6

Estates outsourcing to central factories. Notes This graph depicts the share of outsourcing estates and their share in acreage sugar planted. H is the hurricane damage index described in Sect. 3. The underlying data comes from the Handbook of Jamaica (1880)

4.3 Impact on first time adoption of modern processing technologies

The role of hurricane destruction in terms of the first time adoption of modern processing technologies is estimated via a set of probit models as follows:

$$\begin{aligned} P\Bigl ( ft{\textbf{PT}}_{it} = 1 \Bigr ) = \Phi \biggl ( \alpha + \beta _{H} H_{it-1} + \beta _{{{\textbf{X}}}} {{\textbf{X}}}_{it-1} + \beta _{{{\textbf{Z}}}} {{\textbf{Z}}}_{i} + \gamma _K + \lambda _t + \epsilon _{it} \biggr ) \end{aligned}$$

(3)

where $ft{\textbf{PT}}$ is a vector of indicator variables indicating a first time adoption in the centrifugal (ftC), the vacuum pan (ftV), or both ($ftC \times ftV$), where the latter may be the simultaneous adoption of both at the same time, or one of these having already adopted the other. We include the same set of controls in Eq. (3) as in Eq. (2), calculated standard errors clustered at the estate level and report the coefficients as marginal effects. One should note that once an estate adopts a technology (or joint technologies) it falls out of the sample. Similarly, estates that already had employed both modern processing technologies at the beginning of our sample would not be included in the estimation. Finally, in our data set we observed estates that adopted both technologies at the same time, as well as sequentially, where in terms of the latter sometimes vacuum pans preceded centrifugals and sometimes it was in the reverse. These complexities meant that we were relatively limited in employing other types of choice models. More specifically, we were limited to modelling the choice of a adoption of either technology or the simultaneous adoption of both, where the latter could be either the simultaneous adoption of both at the same time, or one of these having already adopted the other. One may also want to note that essentially all estates by the beginning of our sample period in 1881 had adopted the steam mill this aspect of technology will not be part of our analysis.

Results for Eq. (3) are given in Table 4. Accordingly, damages to an estate from a hurricane does not induce it to adopt a centrifugal for its the separation of molasses from the sugar in sugar production. Allowing for a lagged response similarly does not suggest that hurricanes can induce evaporation process upgrading.²⁹ One may also note that all controls are insignificant predictors of employing a vacuum pan. In contrast, an estate is significantly more likely to adopt a vacuum pan for evaporation once it has been damaged by a hurricane, at least in the year of the event. The estimate on the coefficient suggests that the average damage once damage occurs would increase the probability of adoption by 5.1 per cent, while the largest observed damage during out sample period induces the likelihood by 11.8 per cent. Older plants are also less likely to adopt a vacuum pan, although at a decreasing rate. The probability joint adoption, either at the same time or conditional on one being adopted, however, is not determined by hurricane damage. Age, in contrast, reduced joint adoption, but again at a decreasing rate.

Table 4 Impact of hurricane on process upgrading

4.4 Hurricanes, first time adoption of modern processing technologies, and access to finance

As noted earlier, purchasing more modern processing technologies may have required significant financing both from a theoretical perspective, and in particular self-financing in our context. More specifically, once there was damage to equipment from a hurricane it may only be when sugar prices had previously been high that self-financing of such an investment was possible for many estates. At the same time, one might also expect that some estates might consider instead a switch to the lower capital intensity production of bananas. To capture the two competing financing aspects of the sugar and banana markets and their role after hurricane damage, we first generate the moving averages of the price of sugar, $\sum _{j=0}^{J}\frac{PS_{i,t-1-j}}{J+1}$) and the price of bananas, $\sum _{j=0}^{J}\frac{PB_{i,t-1-j}}{J+1}$), where J and then interacted these with $H_{it-1}$ in Eq. (3). One should note that since we continue to include time dummies in the regression we do not need to include the prices themselves as their effects are captured already through the time dummies (which also capture all other common time varying factors as well). Assuming that our geographic controls, i.e. latitude, longitude, and the pre-sample period hurricane damage distributional parameters, allow us to interpret the coefficient on H causally, then the identifying assumption for doing the same for its interaction with the sugar and bananas moving average prices is that there were no other time varying factors correlated with these that might have affected how hurricane damage impacted first time adoption of modern processing technologies. We are not aware of any potential candidates for such a violation.

The results of re-estimating Eq. (3) with the hurricane damage price interaction terms are shown in Table 5, where we for each first time adoption variable experiment with up to 3 years moving average definitions of prices. Examining the first time adoption of centrifugals in the first four columns shows that the upgrading effect of hurricanes does not depend on the price of sugar no matter how long (up to the previous 3 years) one allows for its impact. In contrast, when one defines $J=1$, i.e. takes into account anamolies of the price of bananas over the last 2 years, the banana price interaction term is significantly negative. This suggests that once an estate is damaged by a hurricane if the price of bananas is high then the incentive to update the separation process technology is lower. Thus, the hurricane induced adoption works is solely dependent on the price of this alternative crop. Intuitively, this may be because a higher price in bananas makes such a long-term and costly investment less attractive as it may be more profitable to switch to bananas cultivation. Using the point estimates suggests that for average hurricane damage, the average observed anamoly (0.05) in the price of bananas relative to its sample mean reduces the probability of employing a centrifugal by 74 per cent.

Table 5 Impact of hurricane on process upgrading and prices

In terms of the upgrading of the evaporation technology, one similarly finds that there is only an effect of hurricanes through the price of bananas, and not sugar. For this, the significant interaction term is for $J=2$, i.e. dependent on a slightly longer horizon of price development of 3 years. Its point estimate is, however, larger, suggesting that an average damaged estate is very unlikely to upgrade to a vacuum pan when the 3 year average price of bananas was at least as large at its average anomaly.

Both the price of sugar and bananas matters when one considers price movements over the longer term of 3 years ($J=2$) for a complete process upgrading. Noticeably in this regard is not only that the estimated coefficient on the price of sugar interaction with H is positive, but also multiple times larger than that of bananas. Thus, sugar estates are particularly sensitive to the price of sugar after a damaging hurricane in terms of considering an upgrade. While the estimated coefficient on the interaction term with the price of bananas lies between that found for the centrifugal and vacuum pan only estimates, the point estimate on that of sugar indicates an unfeasibly (over 100 per cent) large probability response to increases in the price of sugar. This supports the general view of the general extreme own price sensitivity of technological investment in the Jamaican sugar industry (Sheridan 1989), but within the context after hurricane damages were incurred. One may also note that the sensitivity of complete upgrading to hurricane damage and prices can be explained by the combined operation of these technologies, as they provide the most efficiency gains, as noted earlier.

As noted in Sect. 2.4.3, there was little private financing available for sugar estates to upgrade their technology during our sample period. The government did on two occasions provide loans to estates in response to hurricanes, and these funds may have been used to upgrade damaged, or perhaps even non-damaged, processing technology. In this regard, as noted earlier, the loan amounts available after the Hurricane of 1903 were very small and certainly not enough to finance such upgrading.³⁰ In contrast, the amount of funds available after the hurricane in 1912 and provided by local loan banks was potentially much larger. In Table 6, we investigate whether these loans may have played a role in technological upgrading by including the estate-level proxy (LOAN) and its interaction term with our hurricane damage index. One should note in this regard that as there were no upgrades to centrifugals in this year, we are unable to examine such upgrading on its own, but rather only in terms of how already having a centrifugal affects upgrading to a vacuum pan. Our identifying assumption is that there are no other time varying estate specific factors that we do not control for that are correlated with the lending of their nearest agricultural societies’ bank branch.

Table 6 Impact of hurricane on process upgrading and financing

The first two columns in Table 6 re-estimate columns three and four of Table 4 but including the 1912 hurricane loan proxy, where we, as with the price specifications, include our geographic controls, parish dummies, and cluster at the estate level. As can be seen, for vacuum pan upgrading while H continues to have a positive significant impact, the coefficient on LOAN is negative and significant. One possible reason may be that when considering vacuum pan upgrading any loans obtained were instead used for repairs. Allowing for the effect of the latter to depend on hurricane damage, while not changing the significant effect of H, both the loan variable and its interaction term become insignificant (see Column 3). Including the price interaction terms (Column 5) does not alter these findings.

For a complete upgrading of processing technology (Column 2), LOAN similarly display a negative impact when included on its own. Including the interaction term with H in the fourth column, however, changes the coefficient on LOAN from significantly negative to significantly positive. Thus, the direct effect of greater local bank lending increases the probability of adopting the newest processing technology. Moreover, the interaction term between the loan variable and the hurricane index is significantly negative. Thus, while the average loan amount per estate extended by the nearest local loan bank increases the likelihood of using modern processing technology, the higher the hurricane damage, the lower this effect would be. This is robust to including the price variables, as shown in the last column. These results suggest that while the loans were used to upgrade, this was less so the greater the hurricane damage incurred. Possibly this may be because hurricanes caused a loss in production, as shown in Sect. 4.1, some of the loan funds had to be used to cover variable costs or cover other financial obligations, and/or were used for the repair or replacement of other production infrastructure. For the average loan amount (£298), the overall implied net impact on the probability of upgrading would still be negative (6.9 per cent) for the average level of hurricane damage. It is only when hurricane damage is rather small, i.e. less than the around 22nd percentile of observed nonzero damage distribution, that there is an overall positive effect on upgrading. Finally, one may want to note that the fact that LOAN is a significant predictor independent of hurricane damage is suggestive of the possibility that loan provision may not have been completely dependent on hurricane damage and/or that our proxy does not capture all hurricane damages relevant for loan provision.

4.5 Robustness checks

We subject our preferred specification of technology adoption of the last column of Table 6 to a number of robustness checks. Firstly, as noted earlier, our identifying assumption is that there were no other common time varying events that may have coincided with sugar price movements that may have also altered the relationship between hurricane damage and process upgrading. One possible culprit might be occurrence of World War I (WWI). In particular, the price of sugar increased substantially during WWI due to the fall in beetroot sugar production (Poggi 1930). However, trade disruptions during WWI may have also reduced the imports of the sugar processing equipment and thus the ability of plantations to upgrade. To investigate the role of these factors, we created, as in Dye (1994), a dummy for WWI and interacted this with the hurricane destruction index ($\hbox {WAR}_{t} \times H_{it}$) as well as the sugar price interaction term with the destruction index ($\hbox {WAR}_{t} \times \sum _{j=0}^{2}\frac{PS_{i,t-1-j}}{3} \times H_{i,t}$). As can be seen from the first column in Table 7, these additional interaction terms are not significant and do not noticeably change the results on the impact of hurricane damages and its interaction terms with loans and sugar and bananas prices.

Table 7 Impact of hurricane on process upgrading ($ftC\times ftV$) and financing: robustness checks

Another common time varying change that might undermine our identifying assumption are sugar duties. In particular, after several decades of zero tariffs the British empire (re-) introduced a sugar duty of 4 shillings and 2 pence in 1901, further changing this to 1 shilling and 10 pence, 14 shilling, 25 shilling and 8 pence, 21 shilling and 4 $\frac{2}{3}$ pence, 9 shilling and 8 $\frac{2}{3}$ pence, 7 shilling and 4 $\frac{2}{3}$ pence, and 4 shilling and 9 $\frac{2}{3}$ pence per CWT in 1908, 1915, 1916, 1919, 1924, 1925, and 1928, respectively. Interacting the deflated (to 1880 prices) value of the sugar duty with the damage index ($\hbox {DUTY}_{t} \times H_{it}$) and with the interaction term of the price of sugar and the damage index ($\hbox {DUTY}_{t} \times \sum _{j=0}^{2}\frac{PS_{i,t-1-j}}{3} \times H_{i,t}$) and including this in the specification does, however, not alter the original findings concerning financing and hurricane damage in any noticeable way; see the second column of Table 6.

Another concern is whether the standard errors on our coefficients are appropriately estimated, where we have thus far assumed these to be clustered by plantation. As noted by Bertrand et al. (2004) in standard settings where there are clear treatment and control groups then it is important to allow for correlation the error term within the level of the treatment. In our setting, however, our treatment differs not only across units but also across time, as storms hit intermittently and each may affect different plantations. To nevertheless take account of the higher-level treatment, we reran the specification in the last column of Table 6 clustering standard errors by storm and year. As Column 3 of Table 6 shows this, if anything reduces the estimated standard errors, and does not change the conclusion when clustering is undertaken at the plantation level.

It could also be that technological upgrading is spatially correlated, which if not taken account of can lead to poorly estimated standard errors (McMillen 1992). To explore this, we allowed for spatial correlation of the error term of plantations within pre-chosen distance thresholds using the plantations geographic coordinates in the spirit of Conley (1999).³¹ The estimated (standard) coefficients along with their standard errors are shown the last four columns of Table 6 for thresholds of 0, i.e. no spatial correlation, and 20, 50, and 100 km, respectively. Accordingly, allowing for spatial correlation over longer distances only reduces the standard errors compared to allowing for no spatial correlation and does not change our results qualitatively, except that with the 100 km distance threshold there is now a direct impact of hurricane damages on technology adoption.

5 Conclusions

In this paper, we investigated how hurricanes, although intrinsically destructive, may have been a creative force in terms of inducing the technologically laggard Jamaican sugar industry in the late nineteenth and early twentieth centuries to become more cost-efficient. To this end, we assembled an exhaustive sugar estate panel data set that, in addition to sugar production factors and ownership characteristics, contained detailed information in terms of the sugar processing technology employed. Combining this with an estate specific proxy of hurricane damage derived from historical storm tracks and a physical wind exposure model allowed us to examine whether hurricanes played a role in encouraging Jamaican sugar plantations to upgrade to modern machinery.

The findings show that hurricanes not only reduced sugar production, but encouraged some estates to exit the sugar production market. But, importantly, they also encouraged surviving firms to upgrade their processing machinery to technologically superior equipment after being damaged. However, this rather costly upgrading would have been dependent on access to substantial financial means, a challenge for the already generally heavily indebted estates. Moreover, at the time the financial market in Jamaica was largely undeveloped, and thus, private external financing was generally not available to sugar estates. In line with this, we show that upgrading after a destructive hurricane depended heavily on the price of sugar. Moreover, price increases in the main alternative export crop at the time, the banana, discouraged such upgrading even when damage was incurred. The government did on occasion provide some lending to the agricultural sector after hurricanes that feasibly could have also been used for upgrading. However, these loans were likely too insignificant to undertake the large investment in new machinery. As a matter of fact, the evidence for local lending after the destructive hurricane in 1912 indicates that only very slightly damaged estates might have been encouraged to use such financing to upgrade.

More generally, our study demonstrates that while the sometimes hostile natural environment in the British West Indies was unquestionably destructive to local economies, it may at the same time also have been a blessing by acting as a catalyst for surviving plantations to adapt technologically to a changing market. Crucial for the latter to happen was access to adequate financing, a factor that still today appears to play an inhibiting role for developing countries facing environmental disasters (see, for example, McDermott et al. 2014; Zhang and Managi 2020). Finally, whether any benefits arising from historical natural disasters such as hurricanes may have persisted to aid local long-term Caribbean economic development, as has already been demonstrated in US contexts of the 1872 Boston fire (Hornbeck and Keniston 2017) and the 1927 Mississippi flood (Hornbeck and Naidu 2014), is a question that as of date remains unaddressed.