Democratic Backsliding, FDI, and Real Lending Interest Rates: Evidence from Turkey, Venezuela, and El Salvador

Summary

When a government weakens democracy, by controlling the courts, silencing the press or rigging elections, do foreign investors pull their money out and do real lending interest rates rise? This paper reports data from 175 countries between 1990 and 2025 and three case studies: Turkey, Venezuela and El Salvador. The worldwide interest-rate measure covers bank lending, not sovereign borrowing costs; government debt is discussed separately in the case studies.

Across all countries, there is no clear pattern: a drop in democracy in one year is usually not linked to a noticeable drop in foreign investment or a rise in interest rates. But the three country stories show that weakening democracy can be very costly, especially when a government also takes control of the central bank or seizes private property. How much it costs seems to depend on what the government breaks, not just how much democracy declines.

1. Introduction

Leaders who weaken democracy often promise that a strong hand will bring stability and growth. If that is wrong, and weakening democracy actually scares away money, then ordinary people pay the price through fewer jobs, a weaker currency and less money for schools and hospitals.

This paper examines foreign investment and real lending interest rates when democratic institutions weaken. Its country case studies also discuss sovereign credit conditions, which are not directly measured by the worldwide regressions.

To answer it, the paper does two things. First, it checks whether the pattern shows up in data from almost every country in the world. Second, it looks closely at three countries where democracy fell sharply, to see what actually happened to their economies.

2. Background

What is democratic backsliding? Political scientist Nancy Bermeo (2016) describes it as a government slowly weakening the institutions that keep democracy working. Today this usually happens step by step, through elected leaders grabbing more power, rather than through a sudden military coup.

Why might investors care? Two ideas explain the link.

Foreign direct investment (FDI) is when a company from one country builds or buys a business in another, such as a factory or a mine. These investments take years to pay off, so companies want fair courts and stable rules that protect them. Jensen (2003) found that democracies tend to attract more FDI for this reason. However, Li and Resnick (2003) point out that democracies may also give foreign companies fewer special deals, so the overall effect is not guaranteed.

Borrowing costs are the interest a government pays when it borrows money. North and Weingast (1989) argued that when a ruler's power is limited, lenders trust that the government will repay, so it can borrow more cheaply. If those limits disappear, lenders may demand higher interest to cover the extra risk. Archer, Biglaiser and DeRouen (2007) found that credit-rating agencies such as S&P, Moody's and Fitch tend to give democracies in developing countries better ratings.

What this paper expects. If these ideas are right, then when democracy weakens, FDI should go down and borrowing costs should go up. The paper also tests a third idea: the damage might be worse when a government attacks property rights or the central bank, and smaller when it keeps its budget under control.

3. Data and Methods

The study combines two public datasets covering 175 countries from 1990 to 2025.

Democracy: the V-Dem project (Coppedge et al., 2026) scores every country's democracy each year from 0 (not democratic) to 1 (fully democratic), based on ratings from thousands of country experts.

Economy: the World Bank (World Bank, 2026) provides yearly FDI, real lending interest rates, economic growth and inflation where available. Indicator FR.INR.RINR is the lending interest rate adjusted for inflation using the GDP deflator; it is not a sovereign bond yield. https://data.worldbank.org/indicator/FR.INR.RINR

The method. Instead of comparing rich countries with poor ones, the study compares each country with itself from one year to the next. It asks: in years when a country's democracy score fell, did its FDI also fall? This avoids mixing up democracy with things that make countries different to begin with, like size or location. The analysis also adjusts for worldwide events such as the 2008 financial crisis, and for each country's growth and inflation.

FDI is measured as a share of the economy (% of GDP). This matters because measuring it in raw dollars lets a few giant economies dominate the results.

Extreme values, such as hyperinflation years or tiny tax havens with huge money flows, are trimmed so that a few unusual cases do not distort the results.

Table 1 shows typical values in the data. Each "country-year" is one country in one year.

The equations. The study estimates two equations, one for each question:

ΔFDIᵢₜ = β₁ · ΔDemocracyᵢₜ + γ · ΔControlsᵢₜ + δₜ + εᵢₜ

ΔInterestRateᵢₜ = β₂ · ΔDemocracyᵢₜ + θ · ΔControlsᵢₜ + δₜ + μᵢₜ

What each part means:

Δ (delta) means "the change from last year." So ΔFDI is how much a country's FDI went up or down compared with the year before.

i stands for the country and t for the year.

ΔDemocracy is the yearly change in the V-Dem democracy score.

β₁ and β₂ (beta) are the numbers the study is trying to find. They measure how strongly FDI and interest rates move when democracy changes.

ΔControls are the yearly changes in economic growth and inflation, so the study does not blame democracy for what the economy was doing anyway.

δₜ (delta-t) adjusts for things that hit every country in the same year, like the 2008 crisis or COVID-19.

ε and μ are the "error" terms: everything else the equation does not capture.

Hypotheses. If the theory is right, β₁ should be positive (when democracy falls, FDI falls too) and β₂ should be negative (when democracy falls, interest rates rise).

Two models. The study runs each equation in two ways, to see whether the answer holds up when the setup gets more careful:

Model 1 (basic): FDI measured in billions of dollars, no controls or year adjustments, and only the 132 countries that have both FDI and interest-rate data.

Model 2 (full): FDI measured as a share of the economy, all countries with data, the controls and year adjustments shown in the equations, extreme values trimmed, and errors grouped by country (because years from the same country are not independent). Model 2 is the study's main model.

To test the third idea, Model 2 is also run with two of V-Dem's more specific scores in place of the overall democracy score: one for property rights and one for how much the courts can limit the government.

4. What the Worldwide Data Shows

Across all 175 countries, falling democracy does not reliably go along with falling FDI or rising interest rates.

Model 1, the basic version, reports a link: in years when democracy fell, FDI measured in dollars tended to fall too. In Model 2, which measures FDI as a share of each economy and adjusts for growth, inflation and worldwide events, that association is not statistically significant. The models use different measures, samples and adjustments, so this comparison alone does not establish why the result changes.

Table 3 reports the estimates. β is the coefficient from the equations. A p-value describes how compatible the observed result, or a more extreme one, is with a zero-effect null hypothesis under the model's assumptions. A value below 0.05 is a conventional significance threshold, not proof of a real or causal effect.

Only one of the specific scores shows a real pattern, and it points the wrong way: when court independence fell, real interest rates tended to fall, not rise. One possible reason is that governments that control the courts also tend to hold interest rates below inflation, but this study cannot test that.

To put a number on it: even in a year when a country's democracy score fell sharply (by 0.05, which happened only about 180 times in the data), Model 2 expects FDI to change by less than a quarter of one percent of the economy, either up or down. In other words, there is no sign of a large, quick, across-the-board penalty.

That does not prove weakening democracy is harmless. Yearly averages across very different countries can hide big effects in particular places, which is why the next section looks at three of them.

5. Three Country Stories

Turkey, Venezuela and El Salvador saw declines in democracy scores but different economic outcomes. The reported before-and-after comparisons are summarized in Table 4.

Turkey: attacking the central bank

A central bank sets interest rates and protects the value of a country's money, so investors want it to be independent from politicians. In March 2021, President Erdoğan fired the head of Turkey's central bank just two days after he raised interest rates to fight inflation. It was the third time since mid-2019 that Erdoğan had fired a central bank chief, and the Turkish lira lost about 15% of its value when markets opened (Al Jazeera). In March 2025, after the arrest of Istanbul's popular opposition mayor, the lira fell as much as 12% to a record low, and the central bank had to spend tens of billions of dollars to stop the slide (Al-Monitor). According to World Bank data, inflation reached 72% in 2022.

Venezuela: seizing property and defaulting

Venezuela is the most extreme case. Its democracy score fell from 0.73 to 0.18, and by 2007 the government had taken control of companies in oil, electricity and telecommunications (IPS). FDI fell by more than half, and in five of the nine years from 2017 to 2025 more foreign investment left the country than came in. In 2017 Venezuela stopped paying its debts (Fortune). Analysts now estimate it owes $150–200 billion in total (Reuters via U.S. News). Falling oil prices and U.S. sanctions also played a big role, so not all of the damage came from the loss of democracy.

El Salvador: punished, then forgiven

In May 2021, the Legislative Assembly, controlled by President Bukele's party, removed all five judges of the Supreme Court's constitutional chamber and replaced them the same day (Global Voices). Four months later, the new court ruled that Bukele could run for a second term in a row, even though the constitution banned it (ConstitutionNet). Foreign investment held up on average (2.1% of the economy from 2019 onward, compared with 1.5% in the ten years before), although it dropped almost to zero in 2022. Lenders, however, grew nervous: in 2022, the rating agency Fitch rated El Salvador's debt "CC," meaning it thought a default was likely, because of a cash shortage, a large debt payment coming due, and the government's unpredictable decisions, such as making Bitcoin legal money (The Block). After the government cut its budget deficit and agreed to a $1.4 billion loan program with the International Monetary Fund, Fitch raised its rating to "B-" in January 2025 (Reuters via Investing.com). That is still a risky rating, but far better than before.

What the stories show. Weakening democracy was costly when the government also damaged the central bank (Turkey) or seized property (Venezuela). In El Salvador, taking over the courts did not cause a lasting drop in foreign investment, and lenders came back once the government got its budget under control. The type of damage seems to matter more than the size of the drop in democracy.

6. Limits of This Study

Interest rates are an imperfect measure. The World Bank's real interest rate is what banks charge borrowers, not what governments pay on their debt. It is also missing for Turkey and El Salvador.

Correlation is not causation. The study shows what tends to happen together, not what causes what. For example, a country in an economic crisis might be more likely to elect a strongman, rather than the other way around.

Other events get mixed in. Oil prices, sanctions and the COVID-19 pandemic all affected these economies at the same time.

Three stories are not a rule. The case studies suggest ideas, but three countries are not enough to prove a pattern for the whole world.

7. Conclusion

The worldwide regressions reported here do not establish a statistically significant association between the overall democracy measure and FDI or real lending interest rates in the main model. They do not directly test sovereign borrowing costs. The case studies suggest possible economic risks when governments undermine central-bank independence, property rights, or debt repayment, but do not establish causation.

Venezuela shows how bad it can get, Turkey shows how quickly markets react when a central bank loses its independence, and El Salvador shows that markets can forgive a government that fixes its finances, even while democracy keeps declining. The lesson for citizens is that a calm market today does not mean weakening democracy is free. The bill may simply arrive later, or through a different door.

Note: The analysis code was produced with help from Claude (Anthropic). All results were checked against the original V-Dem and World Bank data.

Appendix: Python Code

The supplied analysis code loads the V-Dem and World Bank files, calculates year-to-year changes, and runs the regression specifications and median summaries. Independent reproduction of the displayed results remains pending; updated downloads may differ from the original data snapshot. It needs Python with the pandas and statsmodels libraries, and the data files saved in the same folder.

"""Democratic backsliding, FDI and interest rates, 1990-2025.

Put these files in the same folder and run:  python replication.py
  - V-Dem-CY-Core-v16.csv  (V-Dem democracy data)
  - World Bank downloads (API_*.csv and Metadata_Country_*.csv) for:
    BX.KLT.DINV.CD.WD, BX.KLT.DINV.WD.GD.ZS, FR.INR.RINR,
    NY.GDP.MKTP.KD.ZG, FP.CPI.TOTL.ZG
"""
import glob, re
import pandas as pd
import statsmodels.formula.api as smf


def find(prefix, code):
    # World Bank file names use dots or underscores in the indicator code
    pat = code.replace('.', '[._]')
    return [f for f in glob.glob(f'{prefix}*.csv') if re.search(pat, f)][0]


def wb(code, name):
    # Load one World Bank file and reshape it to one row per country-year
    d = pd.read_csv(find('API_', code), skiprows=4)
    years = [str(y) for y in range(1990, 2026)]
    return (d.melt(id_vars=['Country Code'], value_vars=years, var_name='Year', value_name=name)
             .astype({'Year': int}))


# 1. Load the democracy scores and merge in the World Bank data
vdem_cols = {'country_text_id': 'cc', 'year': 'Year', 'v2x_polyarchy': 'Dem',
             'v2xcl_prpty': 'Property', 'v2x_jucon': 'Courts'}
vd = pd.read_csv('V-Dem-CY-Core-v16.csv', usecols=list(vdem_cols)).rename(columns=vdem_cols)
vd = vd[vd.Year >= 1990]
meta = pd.read_csv(find('Metadata_Country_', 'BX.KLT.DINV.CD.WD'))
countries = set(meta.loc[meta['Region'].notna(), 'Country Code'])   # real countries, not regions

df = vd
for code, name in [('BX.KLT.DINV.CD.WD', 'FDI_USD'), ('BX.KLT.DINV.WD.GD.ZS', 'FDI_GDP'),
                   ('FR.INR.RINR', 'RIR'), ('NY.GDP.MKTP.KD.ZG', 'GDPG'), ('FP.CPI.TOTL.ZG', 'CPI')]:
    df = df.merge(wb(code, name).rename(columns={'Country Code': 'cc'}), on=['cc', 'Year'], how='inner')
df = df[df.cc.isin(countries)].sort_values(['cc', 'Year']).reset_index(drop=True)
print(f'Panel: {len(df):,} country-years, {df.cc.nunique()} countries')
print(df[['Dem', 'FDI_GDP', 'RIR', 'GDPG', 'CPI']].median().round(2).to_string())   # typical values

# 2. Year-to-year changes (the "delta" in the equations), only between back-to-back years
gap = df.groupby('cc')['Year'].diff()
for c in ['Dem', 'Property', 'Courts', 'FDI_USD', 'FDI_GDP', 'RIR', 'GDPG', 'CPI']:
    df['d_' + c] = df.groupby('cc')[c].diff().where(gap == 1)
df['d_FDI_B'] = df['d_FDI_USD'] / 1e9                 # dollars -> billions

# 3. Trim extreme values (top and bottom 1%)
for c in ['d_FDI_GDP', 'd_RIR', 'd_GDPG', 'd_CPI']:
    lo, hi = df[c].quantile([.01, .99])
    df[c + '_w'] = df[c].clip(lo, hi)


# 4. Run the regressions
def fit(formula, data, cluster=True):
    used = set(re.split(r'[\s~+()]+', formula))
    d = data.dropna(subset=[v for v in data.columns if v in used and v != 'Year'])
    if cluster:   # errors grouped by country
        m = smf.ols(formula, d).fit(cov_type='cluster', cov_kwds={'groups': d['cc']})
    else:
        m = smf.ols(formula, d).fit(cov_type='HC1')
    return m, d.cc.nunique()


controls = ' + d_GDPG_w + d_CPI_w + C(Year)'   # growth, inflation, year effects
basic = df.dropna(subset=['RIR', 'FDI_USD'])    # Model 1 sample: countries with both FDI and rate data
runs = [
    ('Model 1, FDI ($ billions)',   'd_FDI_B ~ d_Dem',                basic, False, 'd_Dem'),
    ('Model 2, FDI (% of GDP)',     'd_FDI_GDP_w ~ d_Dem' + controls, df,    True,  'd_Dem'),
    ('Model 1, real interest rate', 'd_RIR ~ d_Dem',                  basic, False, 'd_Dem'),
    ('Model 2, real interest rate', 'd_RIR_w ~ d_Dem' + controls,     df,    True,  'd_Dem'),
    # Third idea: do property rights or court independence matter more? (Model 2 setup)
    ('Property rights -> FDI',      'd_FDI_GDP_w ~ d_Property' + controls, df, True, 'd_Property'),
    ('Property rights -> rate',     'd_RIR_w ~ d_Property' + controls,     df, True, 'd_Property'),
    ('Court independence -> FDI',   'd_FDI_GDP_w ~ d_Courts' + controls,   df, True, 'd_Courts'),
    ('Court independence -> rate',  'd_RIR_w ~ d_Courts' + controls,       df, True, 'd_Courts'),
]
for label, formula, data, cluster, x in runs:
    m, n_countries = fit(formula, data, cluster)
    print(f'{label:30s} beta = {m.params[x]:6.2f}   p-value = {m.pvalues[x]:.3f}   '
          f'N = {int(m.nobs):,}   countries = {n_countries}')

Sources

Research

Archer, C., Biglaiser, G., & DeRouen, K. (2007). Sovereign bonds and the "democratic advantage": Does regime type affect credit rating agency ratings in the developing world? International Organization, 61(2), 341–365.

Bermeo, N. (2016). On democratic backsliding. Journal of Democracy, 27(1), 5–19.

Jensen, N. M. (2003). Democratic governance and multinational corporations: Political regimes and inflows of foreign direct investment. International Organization, 57(3), 587–616.

Li, Q., & Resnick, A. (2003). Reversal of fortunes: Democratic institutions and foreign direct investment inflows to developing countries. International Organization, 57(1), 175–211.

North, D. C., & Weingast, B. R. (1989). Constitutions and commitment. Journal of Economic History, 49(4), 803–832.

Data

Coppedge, M., Gerring, J., Knutsen, C. H., Lindberg, S. I., Teorell, J., et al. (2026). V-Dem [Country-Year/Country-Date] Dataset v16. Varieties of Democracy (V-Dem) Project. https://doi.org/10.23696/vdemds26

World Bank. (2026). World Development Indicators: FDI, net inflows (current US$ and % of GDP); Real interest rate; GDP growth; Inflation, consumer prices. https://data.worldbank.org/indicator

News

Al Jazeera. (2021, March 22). Turkish lira plunges after Erdogan sacks central bank chief. https://www.aljazeera.com/economy/2021/3/22/turkish-lira-plunges-after-erdogan-sacks-central-bank-chief

Al-Monitor. (2025, April 17). Turkey hikes interest rate to 46% amid turmoil over Imamoglu's arrest. https://www.al-monitor.com/originals/2025/04/turkey-hikes-interest-rate-46-amid-turmoil-over-imamoglus-arrest

ConstitutionNet. (2021, September 6). In El Salvador, supreme court rules presidents can serve two consecutive terms. https://constitutionnet.org/news/el-salvador-supreme-court-rules-presidents-can-serve-two-consecutive-terms

Fortune. (2017, November 21). Venezuelan bondholders are preparing for a fight if the country defaults on its $60 billion debt. https://fortune.com/2017/11/21/venezuela-debt-restructure-maduro-default

Global Voices. (2021, May 5). El Salvador's parliament removes checks and balances on President Bukele's powers. https://globalvoices.org/2021/05/05/el-salvadors-parliament-removes-checks-and-balances-on-president-bukeles-powers/

Investing.com / Reuters. (2025, January 7). Fitch upgrades El Salvador after IMF deal eases financing needs. https://investing.com/news/economic-indicators/fitch-upgrades-el-salvador-after-imf-deal-eases-financing-needs-3801493

IPS. (2007, May). Economy-Venezuela: Nationalisation drive takes broad aim. https://www.ipsnews.net/2007/05/economy-venezuela-nationalisation-drive-takes-broad-aim/

The Block. (2022, September 15). Fitch downgrades El Salvador's rating to 'CC'. https://www.theblock.co/news/markets/2022-09-15-fitch-downgrades-el-salvadors-rating-to-cc-170625

U.S. News / Reuters. (2026, September 23). Trump, Venezuela's Rodriguez discuss debt overhaul, Rubio says. https://www.usnews.com/news/world/articles/2026-09-23/trump-venezuelas-rodriguez-discuss-debt-restructuring-rubio-says