Climatic signatures in crops and grain prices in 19th-century Sweden

Introduction

Ongoing debate about the vulnerability of future societies to climate change is increasingly focusing on growth and development of the global economy (Intergovernmental Panel on Climate Change (IPCC), 2007; Stern, 2007). Economic estimates of the impact of climate change are mainly based on ‘damage functions’ that relate gross domestic product losses to increases in temperature. Market impacts include effects on climate- and weather-sensitive sectors such as agriculture, forestry, fisheries, and tourism; damage to coastal areas from sea-level rise; changes in energy expenditures (for heating or cooling); and changes in water resources.

In pre-industrial European societies, the growth and fluctuations of harvests had noticeable influence on the growth and fluctuations of the overall economy (Edvinsson, 2009a, 2009b). One or more consecutive harvest failures had catastrophic consequences, as reserves would not usually be sufficient to last more than one year. However, over the last few hundred years, the focus of agricultural activity has shifted from providing food to minimise the risks of starvation, to maximisation of profits (Edvinsson, 2009a; Leijonhufvud, 2001).

A key question to be studied is the influence of prevailing weather and climate on crop yields and prices. The relationships between climate/weather, crop yields and price fluctuations have attracted the interest of economists studying empirical business cycles since the turn of the last century. Well-known contributors have been William S Jevons (1884), Henry L Moore (1914, 1923) and William H Beveridge (1921, 1922). More recently, the science of economic history has focused on the influence of climatic changes on important economic and demographic variables (e.g. Bauernfeind and Woitek, 1999; de la Croix et al., 2009; Edvinsson, 2005, 2009a, 2009b; Jutikkala, 1975; Le Roy Ladurie, 1971; Pfister, 1988; Utterström, 1988). This paper investigates the relationships between prevailing temperatures, crop yields and domestic agricultural prices in Sweden over the 19th century and the beginning of the 20th century, the period of rising influence of globalisation in the study region (e.g. Bohlin, 2010). In doing so, we aim to help improve wider understanding of the effects of climate on human ecology in the context of an agricultural society undergoing industrialisation. This requires data on environmental fluctuations with high chronological precision. Adjacent to Sweden, Finnish documentation of historical crop yields, agrophenology and early meteorological observations have already enabled assemblage of overlapping time-series of preindustrial crop yields and agricultural activities in the context of the high-latitude climate of northern Europe (Holopainen, 2004, 2006; Holopainen and Helama, 2009; Holopainen et al., 2006, 2009; Linkola, 1924; Rickard et al., 2010; Tornberg, 1989). These previous studies have shown that documentary evidence of historical agriculture and past climate variability can be used to statistically quantify the interrelationships of environment and man, and that the prevailing weather has indeed been a great modulator of agricultural success and thus nutritional conditions during the past centuries (Holopainen and Helama, 2009; Rickard et al., 2010). The evidence of climatic signatures in ecological processes has been shown in numerous systems (Forchhammer and Post, 2000). Aided by the abundance of relevant historical data in Sweden (Jörberg, 1972a, 1972b; Moberg and Bergström, 1997; Ohlsson, 1959), we aimed to investigate the role in this region of prevailing weather and climate variability in influencing agricultural yield, crop price and the relationship between them. Furthermore, the implications of the results are considered in the context of the prevailing climate conditions during the contemporaneous climatic stage of ‘Little Ice Age’ (Grove, 2004). We have previously shown that ambient temperature in this period has strong effects on agricultural yield in pre-industrial Finland (Holopainen and Helama, 2009).

Materials

Two situational factors of note have strong influences on Sweden’s climate. These are its northern latitude (between 55 and 69°N) and the shelter from mild and wet Atlantic winds by the mountains along its western border with Norway. Most of the country has a typical continental climate with a moderate to large temperature range between summer and winter. In this study, the meteorological observations made in Stockholm, Uppsala and Tornedalen were chosen to represent the climatic data. Temperature measurements were available in all three locations from the year 1802 onwards. Calculation and homogenization of monthly temperature means are presented and discussed in detail by Moberg and Bergström (1997) and Klingbjer and Moberg (2003). Here, the data of the three sites were averaged into monthly estimates of Swedish temperatures. It is likely that the monthly mean temperatures simplify the variations in weather and there may be several other factors (e.g. rain, occasional night frost) which greatly impact on harvests. However, such factors exhibit high levels of spatial variation that cannot be easily incorporated into a country-level analysis. On the other hand, monthly mean temperature data exhibit high synchrony at long distances (Holopainen and Helama, 2009; Koenig, 2002).

The data for crop yield fluctuations comes from the tables compiled by the Swedish Central Bureau of Statistics (Ohlsson, 1959). Here we made use of the ‘General Crop Index’ (Ohlsson, 1959: table E12). This is an annual index of the size of the countrywide harvest as a whole based on the data on relative (per-sowing) yields of all crops (Ohlsson, 1959), and is available from 1786 onwards. In addition, we also used the annual absolute yields of specific crops wheat, rye, barley and oats (Ohlsson, 1959: table E18). These data are available from 1860 onward; the time series are presented in Figure 1a. The general crop index ranges between zero and five with an indication of average crop given by the index value of three. The species-specific crops are given for the country in 1×10⁶ kg (‘1000-tal ton’) per year.

OPEN IN VIEWER

Figure 1.

The series of the grain-specific crop and their prices shown for initial values as adapted from literature (a) and after the removal of the long-term component of the variability (b). The series are shown over the common period for the grain-specific crop and price data (1860–1914).

In his book A History of Prices in Sweden 1732–1914, Lennart Jörberg (1972a) assembled and analyzed prices on a large number of agricultural products. The data cover prices of 61 products during 183 years across Sweden. Our study is based on only the grain prices (rye, barley, oats and wheat; see Figure 1a) because of their primary importance to society. The data set is based on ‘market scales’, which were annually established in Swedish counties from the early 18th and well into the 20th century. Market scales represented official prices and were designed to translate tax payments and other government fees in kind into a monetary value. Market scales were negotiated between payers and receivers of tax on the basis of late December market prices in the region’s boroughs, and so did not represent prices from actual contracts. Market price scale assessments were improved as time went by and in 1855 a so-called B-scale was introduced that reflected cash sales prices collected on a monthly basis. As a result of his detailed analysis, Jörberg (1972a, 1972b) concluded that market scale prices are fairly accurate representations of market prices. The data for wheat, rye, barley and oats were used here owing to their availability since 1803. In this study, the series of species-specific prices were estimated as averages of territory-level price series. Subsequently, the series of average grain price was calculated as an arithmetic mean of the four series.

Methods

First, we eliminated the trend and the mid-frequency components of the variability of crop yields and prices, which could bias the climatic comparisons. We applied a cubic smoothing spline with 50-year frequency response (with 50% cut-off) (Cook and Peters, 1981) to all original series (yields and prices). The target high-frequency component of the variability was obtained by subtraction between the initial and modelled (by 50-year spline) series of the data (shown for the four crop species in Figure 1b). While the initial series show the changes of crop yield and price throughout the study period (Figure 1a), the transformed series no longer exhibit trends or mid-frequency variations but shows the seesaw of the positive and negative departures about the long-term average (Figure 1b). The statistical distribution of each series was evaluated (Jarque and Bera, 1987) but no indications for strong deviation from normality were found. Correlations between temperature, crop and price series were compared using Pearson correlations.

The common period for which temperature data, general crop index and the average grain price data are all available is 1803–1914. The detrended General Crop Index and average crop price were correlated to detrended mean temperatures of each of the 14 months prior to the harvest (spanning from August of previous year to September of the harvest year). The rationale of using winter months (i.e. those months preceding the summer of harvest) alongside the summer months was the use of winter varieties. The grains of wheat and rye, the two most valuable species, were sown in the previous autumn (Gadd, 2000) and these grains were thus under the influence of weather throughout the whole winter. The common period for which temperature data and grain-specific crop and price data are available is 1860–1914. As with the General Crop Index and average price, detrended crop-specific yield and price series were correlated to detrended mean temperatures of the 14 months prior to harvest.

These data series enabled comparison of the relationships between yields and prices of all crops, specifically wheat, rye, barley and oats, as well as their seasonally and subseasonally structured responses to temperature. Comparison between such climatic responses was made following Fritts (1982), by plotting the correlations of each month as a function of climate versus price (vertical axis) and climate versus crop (horizontal axis) correlations. In addition to Pearson correlations (e.g. Briffa et al., 2002), the climatic signatures in crop yield and grain price series were examined using principal component regression (Biondi and Waikul, 2004). The comparisons between the correlation and regression coefficients are given in order to show the robustness of the results in the contexts of the statistical methodology, and were each analysed for between cold versus warm-season structured responses to weather.

Results

The General Crop Index was found to correlate negatively with the average grain price (Figure 2a). The finding indicated that a surplus of crop were associated with lower grain prices. Overall, mean monthly temperature was positively correlated with General Crop Index (Figure 3a) and negatively correlated with average grain price (Figure 3b). The association between these structured responses was strongly negative (Figure 3c), such that the monthly temperatures that had the strongest positive influences on crop yield had the strongest negative influences on crop price.

OPEN IN VIEWER

Figure 2.

Comparison between the general crop index and the average grain price (1803–1914) (a). Species-specific crop and price exemplified by the comparison using the data of wheat (1860–1914) (b).

OPEN IN VIEWER

Figure 3.

Structured response (1803–1914) of the general crop index (GCI) to monthly temperatures of previous (lower case letters) and concurrent (capital letters) year to the harvest. Horizontal line shows the level of statistical significance (p<0.05) (a). Structured response 1803–1914) of the average grain price (AGP) to monthly temperatures of previous (lower case letters) and concurrent (capital letters) year to the harvest (b). Comparison between the responses of the GCI and AGP to the monthly temperatures (c).

Next, the above analyses were carried out using the specific data on yield and prices of wheat, rye, barley and oats. The results for wheat exemplify the general pattern: As with the General Crop Index and average grain price structured response, the mean monthly temperatures which correlated positively with crops were in general correlated negatively with price, and vice versa (Figure 4). This negative association of structured responses was remarkably stronger than the association between the actual crop and price of wheat (Figure 2b). That is, the association of structured responses was characterised by strongly negative correlation coefficients (Figure 4c), whereas the association between the actual amount of wheat crop and its historical price was characterised by only slightly negative correlation (Figure 2b). These findings were similar for wheat, rye, barley and oats (Figure 5a). Between-crop correlations in price were stronger than those of yield (Figure 5b).

OPEN IN VIEWER

Figure 4.

Structured response (1860–1914) of the wheat crop to monthly temperatures of previous (lower case letters) and concurrent (capital letters) year to the harvest. Horizontal line shows the level of statistical significance (p<0.05) (a). Structured response of the wheat price to monthly temperatures of previous (lower case letters) and concurrent (capital letters) year to the harvest (b). Comparison between the responses of the wheat crop and price to the monthly temperatures (c).

OPEN IN VIEWER

Figure 5.

Comparison of intraspecific correlations (1860–1914) between the crop and price values (i.e. the actual quantity variations) and between the structured responses of crop and price, as calculated separately for wheat, rye, barley and oats (a). Comparison of interspecific correlations (1860–1914) between the wheat (W), rye (R), barley (B) and oats (O) crops and prices (b). Statistically significant (p<0.05) correlations are presented with an asterisk.

Interestingly, the influence of temperature did not change when the influences of ‘cold season’ (previous August through concurrent February; Figure 6a) and ‘warm season’ (March through September; Figure 6b) months were considered separately. A comparative examination of the correlation (Figure 6a, b) and PCA regression coefficients (Figure 6c, d) did not change the interpretation, as both types of results indicated similarly negative association.

OPEN IN VIEWER

Figure 6.

Comparison between cold-season (preceding August through concurrent February) (a, c) and warm-season (March through September) (b, d) structured responses and between the coefficients from Pearson correlations (a, b) and principal component regressions (c, d). Each plot combines the monthly coefficients of crop and price data as calculated for wheat (see Figure 4c), rye, barley and oats.

Discussion and conclusions

In this study, we present new evidence on the interrelationships between climatic fluctuations, grain crops and prices. We have focused on the case of Sweden because of the continual availability of climatological, agricultural and economic data for a common period from the early 19th century onwards (Jörberg, 1972a, 1972b; Moberg and Bergström, 1997; Ohlsson, 1959). Notably, we found consistent evidence for positive correlations between the mean monthly temperatures and crop yields. The warmness of the prevailing weather and climate would apparently be beneficial for crop yields, at least in the study region and period. These results complement those obtained previously from adjacent areas in southwest Finland where the temporal variations in historical grain figures (ratio between sown and harvested grain) were shown to correlate positively with the temperatures of growing season (Holopainen and Helama, 2009). Using the agrophenolocial data of rye (phenological records of other species were not available), we found that early ripening was associated with higher crop yield (Holopainen and Helama, 2009). It could be reasonably hypothesised that the relationship between temperature and crop yield is mediated by early ripening.

In a global perspective, the study region is situated in the agricultural periphery of northern Europe. This is a phytogeographical position whose climate entails significant agricultural hardships (Parry, 1975; Parry and Carter, 1985). We have suggested (Holopainen and Helama, 2009) that these hardships were amplified through increased coolness of climate conditions over this study period, because of its overlapping with the ‘Little Ice Age’ (Grove, 2004). According to a recent review by Matthews and Briffa (2005), the ‘Little Ice Age’ interval can be characterised by a significant drop of Northern Hemisphere temperatures especially between the years 1570 and 1900. Essentially, the purpose of agricultural activity at this time was to minimise the risk of crop failure in order to prevent starvation (Edvinsson, 2009a; Leijonhufvud, 2001). The results shown here quantify the extent to which this risk could be influenced by climate and weather.

As would be expected from the positive relationships between weather and crop yields, the associations between weather and prices are predominantly negative. Moreover, the associations between the actual crop quantities and prices were negative. The cooler the prevailing weather, the higher grain prices were observed. Further analysis showed that the associations between the structured responses of crops and prices were strongly negative, potentially indicating a substantial role for ambient temperature in mediating the relationship between yield and price. These results were consistent whether calculated using the general crop index or separately using the intraspecific values for wheat, rye, barley or oats.

Several economic historians have investigated the influence of past climate changes on economic and demographic variables (e.g. Bauernfeind and Woitek, 1999; Jutikkala, 1975; Le Roy Ladurie, 1971; Pfister, 1988; Pfister and Brázdil, 1999). Our study lends support to the idea that climatic and economic fluctuations are related to one another, at least when crop yield is a strong determinant of price. Hence, we have thrown light on the historical price fluctuations of grains in the context of temperature perturbations, from the Swedish perspective.

Crop prices in successive years were not independent, as proved mathematically by Jörberg (1972b: 386) who found that the price series were governed by a significant positive autocorrelation, particularly at a lag of one year. That is, an increase in grain demand could result in increased price but the economic response for altered supply may not be of single-year duration. Given the role of historical agriculture in minimising the risks of starvation (Edvinsson, 2009a; Leijonhufvud, 2001), it is reasonable to suspect that societies aimed at acting as a compensator of price extremes. That the interspecific correlations of prices were indeed higher than the correlations between crops could be taken as evidence for this. Our results suggest that the agricultural economy depended on the influence of weather events via effects on the compound success of crops. Deteriorated climate, causing a failure of any one crop species, could thus be expected to result in increased price of all crops over one or more years. Such interdependence could explain the higher correlations between the interspecific prices compared with those between crops.

Apart from climatic factors, we are aware that the grain price fluctuations have been influenced by several other non-climatic factors. In the late nineteenth century, the world economy was beginning to enter an era of globalisation. The emergence of an international grain market started to influence the grain prices widely (Bohlin, 2005, 2010). Interestingly, the influence of the arising international markets might also be detectable in our results. As shown by the black bars in Figure 5a, the correlations between the crop and price values were poorest for wheat and rye, while the highest correlations were obtained for oats. This may not actually be surprising in the context of international trade. The results showed high consistency with the import penetration ratio (which measures the degree to which an economy is exposed to foreign competition) of the domestic market for the same grain species in contemporaneous Sweden: the import penetration was highest for wheat, then for rye and lowest for barley and oats (see Bohlin, 2010: table 1). The wheat economy was integrated internationally, as was rye (although less so because of some tariff protection from 1888 onwards, Bohlin, 2005, 2010). Oats were hardly imported into Sweden at this time (see also Ohlsson, 1972: tables 2.1 and 3.1).

In addition to rising globalisation, it is likely that there have been other non-climatic factors behind the grain price fluctuations. A significant part of the economic and social transformation of the 19th century was increased population growth and mass emigration to the United States (Runblom and Norman, 1976). From 1750 to 1850 the population in Sweden almost doubled from 1,800,000 to 3,500,000. The growth was caused by shifts in age patterns in mortality, in particular a dramatic decline in infant mortality (de la Croix et al., 2009). At the same time there were significant agricultural modernisation and reforms in most of Europe (e.g. Gadd, 2000). From a methodological viewpoint, it is likely that many of these long- and mid-term effects were likely eliminated from the data in our study because of our filtering of the studied time-series prior to correlation analyses.

In conclusion, we have shown that climate–agriculture and climate–economy associations had implications for historical grain crop and price variations in pre-industrial Sweden, not only over restricted catastrophic starvation events but as a pervasive background forcing under the full range of environmental and economic conditions in effect over this period. We suggest that the cooler climate regime of the ‘Little Ice Age’ is likely to have amplified the effects of climate on crop yields. The ongoing anthropogenic warming of climate represents a reversed climate shift and could thus lead to new and unexpected effects on agriculture and economies.