Climate-driven technical change: seasonality and the invention of agriculture

Transcription

1 Climate-driven technical change: seasonality and the invention of agriculture Andrea Matranga Universitat Pompeu Fabra CaSEs seminars February 17, / 31

2 Introduction I focus on two puzzles related to the Neolithic Revolution Two puzzles: Invented independently by different groups, around the same time. Farmers worked more, and ate less. I propose a new story which answers both puzzles, and test the theory empirically. 2 / 31

3 Figure 1 : Locations where agriculture was invented, and their respective dates (years Before Present). 3 / 31

4 Seasonality, storage, and agriculture Agriculture and sedentism: which came first? Seasonality made food scarce in large areas at the same time. Nomads could not find food anywhere within migratory range. They became sedentary in order to store food. Being sedentary made invention of agriculture easier. Transition is optimal for rational population at equilibrium. No change in preferences, fertility, technology, species composition, or transitory shocks. 4 / 31

5 Figure 2 : Right panel: Seasonal locations became more common shortly before agriculture was invented. Left panel: binned scatterplot of temperature seasonality and adoption; early adopters tend to be highly seasonal, and vice versa. 5 / 31

6 k Humans Leave Africa 60K Broad Spectrum Revolution Years Before Present 22k LGM 12k 5k NeolithicPeriod 0 Precession x Eccentricity Insolation at 65N in July Axial tilt Figure 3 : Variation in orbital forcing (black, effects of axial tilt, and the combined effect of precession and eccentricity). Data from Berger (1992). Seasonality conditions at 65 N (red) are indicative of those in the rest of Northern hemisphere. 6 / 31

7 Related Literature Neolithic triggers: gradual population growth (Locay 1989), warmer climates (Diamond 1997), drier climates (Childe 1935), more stable climates (Richerson et al 2001), less stable climates (Morand 2002), intermediately stable climates (Ashraf and Michalopoulos 2014). Loss in consumption per capita: Man s Worst Mistake (Diamond 1987); defense (Seabright and Rowthorn 2010); non-food goods (Weisdorf 2009), social reasons (Acemoglu and Robinson 2012). Long-run effects of Neolithic: gradual spread (Ammerman and Cavalli-Sforza 1971), earlier state formation (Wittfogel 1957), fixed capital (Diamond 1997), greater farming productivity (Harlan 1995), cultural adaptations (Alesina et al. 2013), ethno-linguistic diffusion (Bouckaert et al. 2012). 7 / 31

8 Rest of Presentation Empirics 1. Climate and the Neolithic Seasonality and invention of agriculture Seasonality and spread of agriculture 2. Geographic variety and agricultural adoption 3. Evidence for consumption seasonality across the transition Conclusions and further work 8 / 31

9 Model: overview Basic assumptions: Agents like eating more on average, and dislike eating irregularly. Pure endowment economy. Endowments vary across space and time. Nomadism and storage are mutually exclusive. Population constant (endogenous population growth in the paper). 9 / 31

10 Model predictions Climate seasonality should make invention of agriculture more likely, and its spread faster. Presence of uncorrelated food sources should delay adoption. When transitioning to farming, consumption per capita should drop both in the short run (loss of dispersed food sources) and in the long run (demographic effect). When transitioning to farming, consumption seasonality should decrease substantially. All of these predictions can be tested econometrically. 10 / 31

11 Data sources Global scale: Invention: Puruggannan and Fuller Seven sites that invented agriculture (+17 other domestication sites), and their dates. Adoption: Putterman and Trainor countries and their dates of adoption. Climate: He Lat 96 Lon 22,000 years. Reconstructed temperature and precipitation for each trimester. Collapsed to 44 periods of 500 years. Western Eurasia: Adoption: Pinhasi et al countries and their dates of adoption. Climate: Hijmans et al averages.climate statistics for 10km squares. Others: altitude from SRTM, distance from sea, latitude, Americas dummy. 11 / 31

12 Construction of explanatory variables. TempSeas = max(temp.warmest, 0) max(temp.coldest, 0) Precip.Wettest Precip.Driest PrecipSeas = MeanPrecip. SeasIndex = max(quantile(tempseas), Quantile(PrecipSeas)) 12 / 31

13 Invention of agriculture Was agriculture invented in highly seasonal places? The dataset is structured as a panel of 1036 land cells (3.75 squares) times 44 periods. For each cell, I have temperature and precipitation mean, temperature and precipitation seasonality, and a dummy for whether agriculture was invented in that particular place and time or not. mean sd min max Year Adop Temp. Seas Precip. Seas Temp. Mean Precip. Mean Seas. Index Observations 1036 Table 1 : Summary statistics for the adoption cross-section dataset. 13 / 31

14 Map of climate and invention Geographic distribution of seasonality Pleasant in 8k BP Independent Agri. Seasonal in 21k BP Seasonal in 8k BP Figure 4 : The map shows the global distribution of seasonal locations. Pink cells were already seasonal in 21k BP. Cells that were seasonal in 8,000, are in Red. Dark blue cells were hospitable in 8,000 BP. Locations that were not hospitable in 8,000 BP are omitted. Most of the areas where agriculture was invented had recently become extremely seasonal. 14 / 31

15 I it = α + β 1 TempDiff it + β 2 PrectSeas it + γ[controls] it + ɛ it Dependent variable: adoption dummy (1) (2) (3) (4) (5) Basic Controls Controls2 ModernSeas SI Neol7 Temp. Seas (0.051) (0.063) (0.106) Precip. Seas (0.633) (0.679) (1.339) Seas. Index (4.021) (3.879) Temp. Mean (0.050) (0.125) (0.129) (0.038) (0.149) Precip. Mean (0.216) (0.625) (0.713) (0.301) (0.713) Abs Lat (0.034) (0.088) (0.101) (0.050) (0.065) Temp. Seas. Today (0.207) Precip. Seas. Today (1.265) Seas. Index Today (2.021) Extra Controls No Yes Yes No Yes p N / 31

16 The spread of agriculture Did agriculture spread faster in seasonal locations? We are interested in the adoption behavior of locationsexposed to existing agricultural technology. From the panel dataset, I drop all observations that are either inhospitable, have already adopted agriculture, or are further than 500km from a location that has already adopted. All observations are absent from the dataset until a neighbor adopts agriculture, then have a series of zeros until they adopt themselves, at which point they receive a one, and are then dropped from the dataset. 16 / 31

17 The spread of agriculture 17 / 31

18 The spread of agriculture 18 / 31

19 A it = α + β 1 T it + β 2 P it + γc it + ɛ it Dependent variable: adoption dummy (1) (2) (3) (4) (5) (6) Linear Linear Geog.Cluster LinearSI Logit Logit+ Geog.Cluster LogitSI main Temp. Seas (0.002) (0.003) (0.011) (0.015) Precip. Seas (0.019) (0.029) (0.092) (0.144) Seas. Index (0.096) (0.506) Temp. Mean (0.002) (0.004) (0.003) (0.010) (0.017) (0.012) Precip. Mean (0.008) (0.015) (0.012) (0.038) (0.071) (0.058) Observations Standard errors in parentheses p < 0.1, p < 0.05, p < 0.01 Table 3 : Spread of agriculture. Neolithic frontier locations only. Regression of adoption dummy on climatic variables. Models 1, 2 and 3: Logit with robust s.e. Model 4, 5 and 6. Linear probability model with robust s.e. 19 / 31

20 Altitude variety and early adoption The relationship between seasonality and the Neolithic is robust, significant, and can be observed both at the global and regional scale. I argue that the association is causal, and due to the incentives to store food. However, other channels are possible. Seasonal climate may favor proliferation of plants that are easy to cultivate (large seeded annual grasses). We need a case in which seasonality is present for plants, but not for nomads. I concentrate on Middle Eastern locations where wild cereals were present, and look whether nomads were able to escape by migrating. 20 / 31

21 The datasets, and four example locations 21 / 31

22 Example: Local Areas 22 / 31

23 Example: Altitude Profiles 23 / 31

24 r(5), r(50), and date of adoption 24 / 31

25 Geographic correlation and early adoption Gain in altitude range (m) Altitude variety and agricultural adoption 0 5km 5 50km km km Distance from site Adopted < 10,000 BP Adopted > 9,000 BP Figure 5 : The graph shows the altitude range that settlers could access (0-5km from site), how much extra range they could access if they were nomadic (5-50km), and how much was out of reach even for nomads (50-100km, and km). When the late adopters eventually became settled, they had to abandon on average 1,250m of altitude range. The early adopters faced a lower opportunity cost: on average they lost only 1,000m. 25 / 31

26 Y i = α + β 1 r(5) + β 2 r(50) + γc i + ɛ i (1) Dependant variable: date of adoption (1) (2) (3) (4) (5) <200km <100km Clim. Means r(200) Smooth Meas. r(5) (0.414) (0.496) (0.580) (0.579) r(50) (0.179) (0.221) (0.267) (0.306) r(3:8) (0.597) r(50:100) (0.254) r(200) (0.266) Temp. Seas (114.1) (116.4) (116.1) Precip. Seas (4268.1) (4417.6) (4040.5) Controls No No Yes Yes Yes Observations R Standard errors in parentheses p < 0.1, p < 0.05, p < 0.01 Table 4 : Regression result for range of altitude within a given radius. 26 / 31

27 Consumption Seasonality and the Neolithic Revolution Figure 6 : Example of Harris lines in an Inuit adult. Regular spacing reflects recurring starvation resulting in growth arrest, followed by rapid increase in food intake and catch-up growth, forming a line of denser bone. Such a regular pattern is extremely unlikely to occur due to illnesses. Source: Lobdell(1984) 27 / 31

28 Conclusions Populations exposed to highly seasonal conditions adopted agriculture considerably ahead of those in more stable environments(both greater probability of adoption, and faster spread). The theory generates additional predictions on the health outcomes across the transition, and the local topography of the earliest settlements. These predictions are consistent with the observed data. These findings suggest that global patterns of climatic seasonality 10,000 years ago played a dominant role in determining where farming first appeared, what crops were domesticated, which ethnic and linguistic groups would proliferate, and where the earliest cities would rise. 28 / 31

29 Further Work Predicting locations favorable to independent invention. Other times and periods where seasonality might be important. Storage and armies 29 / 31